From 14ced893e32e0c1c5e36f5e986c9b4013415009d Mon Sep 17 00:00:00 2001
From: Paul Keen <pftg@Pauls-MacBook-Pro.local>
Date: Tue, 7 Jul 2026 22:57:11 -0400
Subject: [PATCH 1/6] feat(apm): add Agent Package Manager configuration for
 reproducible AI tooling

Add apm.yml manifest (jetthoughts/site@0.1.0) as producer package with:
- 5 instructions (project-rules, voice-guide, icp-target, tdd-enforcement, css-consolidation)
- 3 prompts (blog-pipeline, cover-image, research)
- 18 agents across core/content/ruby/validation
- 5 skills (blog-post-pipeline, css-bem-migration, visual-regression-testing, cover-image-generation, content-voice-enforcement)
- 10 MCP servers (claude-context, context7, brave-search, lean-ctx, chrome-devtools, package-search, github-mcp-server, searxng, ollama, milvus)
- 8 APM targets: claude, copilot, codex, cursor, gemini, opencode, windsurf, kiro

Add compose.mcp.yml + Makefile for per-repo Docker MCP services.
Update Brewfile for ollama, add Makefile for mcp-up/mcp-down/mcp-status.
Refactor thin wrappers (GEMINI.md, QWEN.md, AGENTS.md, .junie/AGENTS.md)
to point to .apm/instructions/ as canonical source.
Commit apm.lock.yaml for reproducible installs.
Add APM getting-started section to README.
---
 .agents/skills/blog-post-pipeline/SKILL.md    |  41 ++
 .../skills/content-voice-enforcement/SKILL.md |  37 ++
 .../skills/cover-image-generation/SKILL.md    |  32 ++
 .agents/skills/css-bem-migration/SKILL.md     |  49 +++
 .../skills/visual-regression-testing/SKILL.md |  41 ++
 .apm/agents/build-monitor.agent.md            |  18 +
 .apm/agents/content/content-creator.agent.md  |  20 +
 .apm/agents/content/content-marketer.agent.md |  10 +
 .apm/agents/content/seo-specialist.agent.md   |  21 +
 .../agents/content/tutorial-engineer.agent.md |  10 +
 .apm/agents/core/analyst.agent.md             |  15 +
 .apm/agents/core/coder.agent.md               |  26 ++
 .apm/agents/core/planner.agent.md             |   9 +
 .apm/agents/core/researcher.agent.md          |  10 +
 .apm/agents/core/reviewer.agent.md            |  17 +
 .apm/agents/core/screenshot-guardian.agent.md |  15 +
 .apm/agents/core/tester.agent.md              |  25 ++
 .apm/agents/hugo-expert.agent.md              |  25 ++
 .apm/agents/ruby/ruby-coder.agent.md          |  15 +
 .apm/agents/ruby/ruby-reviewer.agent.md       |  14 +
 .../validation/qa-browser-tester.agent.md     |  16 +
 .../validation/ui-problem-diagnosis.agent.md  |  10 +
 .../validation/ux-browser-validator.agent.md  |  14 +
 .../css-consolidation.instructions.md         |  30 ++
 .apm/instructions/icp-target.instructions.md  |  46 ++
 .../project-rules.instructions.md             |  79 ++++
 .../tdd-enforcement.instructions.md           |  62 +++
 .apm/instructions/voice-guide.instructions.md |  68 +++
 .apm/prompts/blog-pipeline.prompt.md          |  75 ++++
 .apm/prompts/cover-image.prompt.md            |  19 +
 .apm/prompts/research.prompt.md               |  26 ++
 .apm/skills/blog-post-pipeline/SKILL.md       |  41 ++
 .../skills/content-voice-enforcement/SKILL.md |  37 ++
 .apm/skills/cover-image-generation/SKILL.md   |  32 ++
 .apm/skills/css-bem-migration/SKILL.md        |  49 +++
 .../skills/visual-regression-testing/SKILL.md |  41 ++
 .claude/agents/analyst.md                     |  15 +
 .claude/agents/coder.md                       |  26 ++
 .claude/agents/content-marketer.md            |  10 +
 .claude/agents/planner.md                     |   9 +
 .claude/agents/qa-browser-tester.md           |  16 +
 .claude/agents/researcher.md                  |  10 +
 .claude/agents/reviewer.md                    |  17 +
 .claude/agents/ruby-coder.md                  |  15 +
 .claude/agents/ruby-reviewer.md               |  14 +
 .claude/agents/screenshot-guardian.md         |  15 +
 .claude/agents/tester.md                      |  25 ++
 .claude/agents/tutorial-engineer.md           |  10 +
 .claude/agents/ui-problem-diagnosis.md        |  10 +
 .claude/agents/ux-browser-validator.md        |  14 +
 .claude/commands/blog-pipeline.md             |  79 ++++
 .claude/commands/cover-image.md               |  23 +
 .claude/commands/research.md                  |  30 ++
 .claude/rules/css-consolidation.md            |  30 ++
 .claude/rules/icp-target.md                   |  46 ++
 .claude/rules/project-rules.md                |  79 ++++
 .claude/rules/tdd-enforcement.md              |  62 +++
 .claude/rules/voice-guide.md                  |  68 +++
 .claude/skills/blog-post-pipeline/SKILL.md    |  41 ++
 .../skills/content-voice-enforcement/SKILL.md |  37 ++
 .../skills/cover-image-generation/SKILL.md    |  32 ++
 .claude/skills/css-bem-migration/SKILL.md     |  49 +++
 .../skills/visual-regression-testing/SKILL.md |  41 ++
 .codex/agents/analyst.toml                    |   3 +
 .codex/agents/build-monitor.toml              |   3 +
 .codex/agents/coder.toml                      |   3 +
 .codex/agents/content-creator.toml            |   3 +
 .codex/agents/content-marketer.toml           |   3 +
 .codex/agents/hugo-expert.toml                |   3 +
 .codex/agents/planner.toml                    |   3 +
 .codex/agents/qa-browser-tester.toml          |   3 +
 .codex/agents/researcher.toml                 |   3 +
 .codex/agents/reviewer.toml                   |   3 +
 .codex/agents/ruby-coder.toml                 |   3 +
 .codex/agents/ruby-reviewer.toml              |   3 +
 .codex/agents/screenshot-guardian.toml        |   3 +
 .codex/agents/seo-specialist.toml             |   3 +
 .codex/agents/tester.toml                     |   3 +
 .codex/agents/tutorial-engineer.toml          |   3 +
 .codex/agents/ui-problem-diagnosis.toml       |   3 +
 .codex/agents/ux-browser-validator.toml       |   3 +
 .codex/config.toml                            |  50 +++
 .cursor/agents/analyst.md                     |  15 +
 .cursor/agents/build-monitor.md               |  18 +
 .cursor/agents/coder.md                       |  26 ++
 .cursor/agents/content-creator.md             |  20 +
 .cursor/agents/content-marketer.md            |  10 +
 .cursor/agents/hugo-expert.md                 |  25 ++
 .cursor/agents/planner.md                     |   9 +
 .cursor/agents/qa-browser-tester.md           |  16 +
 .cursor/agents/researcher.md                  |  10 +
 .cursor/agents/reviewer.md                    |  17 +
 .cursor/agents/ruby-coder.md                  |  15 +
 .cursor/agents/ruby-reviewer.md               |  14 +
 .cursor/agents/screenshot-guardian.md         |  15 +
 .cursor/agents/seo-specialist.md              |  21 +
 .cursor/agents/tester.md                      |  25 ++
 .cursor/agents/tutorial-engineer.md           |  10 +
 .cursor/agents/ui-problem-diagnosis.md        |  10 +
 .cursor/agents/ux-browser-validator.md        |  14 +
 .cursor/commands/blog-pipeline.md             |  79 ++++
 .cursor/commands/cover-image.md               |  23 +
 .cursor/commands/research.md                  |  30 ++
 .cursor/mcp.json                              |  78 ++++
 .cursor/rules/css-consolidation.mdc           |  34 ++
 .cursor/rules/icp-target.mdc                  |  50 +++
 .cursor/rules/project-rules.mdc               |  83 ++++
 .cursor/rules/tdd-enforcement.mdc             |  66 +++
 .cursor/rules/voice-guide.mdc                 |  72 ++++
 .gemini/commands/blog-pipeline.toml           |   1 +
 .gemini/commands/cover-image.toml             |   1 +
 .gemini/commands/research.toml                |   1 +
 .gemini/settings.json                         |  68 +++
 .github/agents/analyst.agent.md               |  15 +
 .github/agents/build-monitor.agent.md         |  18 +
 .github/agents/coder.agent.md                 |  26 ++
 .github/agents/content-creator.agent.md       |  20 +
 .github/agents/content-marketer.agent.md      |  10 +
 .github/agents/hugo-expert.agent.md           |  25 ++
 .github/agents/planner.agent.md               |   9 +
 .github/agents/qa-browser-tester.agent.md     |  16 +
 .github/agents/researcher.agent.md            |  10 +
 .github/agents/reviewer.agent.md              |  17 +
 .github/agents/ruby-coder.agent.md            |  15 +
 .github/agents/ruby-reviewer.agent.md         |  14 +
 .github/agents/screenshot-guardian.agent.md   |  15 +
 .github/agents/seo-specialist.agent.md        |  21 +
 .github/agents/tester.agent.md                |  25 ++
 .github/agents/tutorial-engineer.agent.md     |  10 +
 .github/agents/ui-problem-diagnosis.agent.md  |  10 +
 .github/agents/ux-browser-validator.agent.md  |  14 +
 .../css-consolidation.instructions.md         |  30 ++
 .../instructions/icp-target.instructions.md   |  46 ++
 .../project-rules.instructions.md             |  79 ++++
 .../tdd-enforcement.instructions.md           |  62 +++
 .../instructions/voice-guide.instructions.md  |  68 +++
 .github/prompts/blog-pipeline.prompt.md       |  75 ++++
 .github/prompts/cover-image.prompt.md         |  19 +
 .github/prompts/research.prompt.md            |  26 ++
 .gitignore                                    |  49 ++-
 .junie/AGENTS.md                              |  26 +-
 .kiro/settings/mcp.json                       |  68 +++
 .kiro/skills/blog-post-pipeline/SKILL.md      |  41 ++
 .../skills/content-voice-enforcement/SKILL.md |  37 ++
 .kiro/skills/cover-image-generation/SKILL.md  |  32 ++
 .kiro/skills/css-bem-migration/SKILL.md       |  49 +++
 .../skills/visual-regression-testing/SKILL.md |  41 ++
 .kiro/steering/css-consolidation.md           |  34 ++
 .kiro/steering/icp-target.md                  |  50 +++
 .kiro/steering/project-rules.md               |  83 ++++
 .kiro/steering/tdd-enforcement.md             |  66 +++
 .kiro/steering/voice-guide.md                 |  72 ++++
 .opencode/agents/analyst.md                   |  15 +
 .opencode/agents/build-monitor.md             |  18 +
 .opencode/agents/coder.md                     |  26 ++
 .opencode/agents/content-creator.md           |  20 +
 .opencode/agents/content-marketer.md          |  10 +
 .opencode/agents/hugo-expert.md               |  25 ++
 .opencode/agents/planner.md                   |   9 +
 .opencode/agents/qa-browser-tester.md         |  16 +
 .opencode/agents/researcher.md                |  10 +
 .opencode/agents/reviewer.md                  |  17 +
 .opencode/agents/ruby-coder.md                |  15 +
 .opencode/agents/ruby-reviewer.md             |  14 +
 .opencode/agents/screenshot-guardian.md       |  15 +
 .opencode/agents/seo-specialist.md            |  21 +
 .opencode/agents/tester.md                    |  25 ++
 .opencode/agents/tutorial-engineer.md         |  10 +
 .opencode/agents/ui-problem-diagnosis.md      |  10 +
 .opencode/agents/ux-browser-validator.md      |  14 +
 .opencode/commands/blog-pipeline.md           |  79 ++++
 .opencode/commands/cover-image.md             |  23 +
 .opencode/commands/research.md                |  30 ++
 .windsurf/rules/css-consolidation.md          |  34 ++
 .windsurf/rules/icp-target.md                 |  50 +++
 .windsurf/rules/project-rules.md              |  83 ++++
 .windsurf/rules/tdd-enforcement.md            |  66 +++
 .windsurf/rules/voice-guide.md                |  72 ++++
 .windsurf/workflows/blog-pipeline.md          |  79 ++++
 .windsurf/workflows/cover-image.md            |  23 +
 .windsurf/workflows/research.md               |  30 ++
 AGENTS.md                                     |  29 +-
 Brewfile                                      |   3 +
 GEMINI.md                                     |  26 +-
 Makefile                                      |  52 +++
 QWEN.md                                       |  30 +-
 README.md                                     |  55 +++
 apm.lock.yaml                                 | 395 ++++++++++++++++++
 apm.yml                                       | 135 ++++++
 compose.mcp.yml                               |  67 +++
 docs/mcp-user-setup.md                        |  40 ++
 opencode.json                                 |  88 ++++
 192 files changed, 5933 insertions(+), 87 deletions(-)
 create mode 100644 .agents/skills/blog-post-pipeline/SKILL.md
 create mode 100644 .agents/skills/content-voice-enforcement/SKILL.md
 create mode 100644 .agents/skills/cover-image-generation/SKILL.md
 create mode 100644 .agents/skills/css-bem-migration/SKILL.md
 create mode 100644 .agents/skills/visual-regression-testing/SKILL.md
 create mode 100644 .apm/agents/build-monitor.agent.md
 create mode 100644 .apm/agents/content/content-creator.agent.md
 create mode 100644 .apm/agents/content/content-marketer.agent.md
 create mode 100644 .apm/agents/content/seo-specialist.agent.md
 create mode 100644 .apm/agents/content/tutorial-engineer.agent.md
 create mode 100644 .apm/agents/core/analyst.agent.md
 create mode 100644 .apm/agents/core/coder.agent.md
 create mode 100644 .apm/agents/core/planner.agent.md
 create mode 100644 .apm/agents/core/researcher.agent.md
 create mode 100644 .apm/agents/core/reviewer.agent.md
 create mode 100644 .apm/agents/core/screenshot-guardian.agent.md
 create mode 100644 .apm/agents/core/tester.agent.md
 create mode 100644 .apm/agents/hugo-expert.agent.md
 create mode 100644 .apm/agents/ruby/ruby-coder.agent.md
 create mode 100644 .apm/agents/ruby/ruby-reviewer.agent.md
 create mode 100644 .apm/agents/validation/qa-browser-tester.agent.md
 create mode 100644 .apm/agents/validation/ui-problem-diagnosis.agent.md
 create mode 100644 .apm/agents/validation/ux-browser-validator.agent.md
 create mode 100644 .apm/instructions/css-consolidation.instructions.md
 create mode 100644 .apm/instructions/icp-target.instructions.md
 create mode 100644 .apm/instructions/project-rules.instructions.md
 create mode 100644 .apm/instructions/tdd-enforcement.instructions.md
 create mode 100644 .apm/instructions/voice-guide.instructions.md
 create mode 100644 .apm/prompts/blog-pipeline.prompt.md
 create mode 100644 .apm/prompts/cover-image.prompt.md
 create mode 100644 .apm/prompts/research.prompt.md
 create mode 100644 .apm/skills/blog-post-pipeline/SKILL.md
 create mode 100644 .apm/skills/content-voice-enforcement/SKILL.md
 create mode 100644 .apm/skills/cover-image-generation/SKILL.md
 create mode 100644 .apm/skills/css-bem-migration/SKILL.md
 create mode 100644 .apm/skills/visual-regression-testing/SKILL.md
 create mode 100644 .claude/agents/analyst.md
 create mode 100644 .claude/agents/coder.md
 create mode 100644 .claude/agents/content-marketer.md
 create mode 100644 .claude/agents/planner.md
 create mode 100644 .claude/agents/qa-browser-tester.md
 create mode 100644 .claude/agents/researcher.md
 create mode 100644 .claude/agents/reviewer.md
 create mode 100644 .claude/agents/ruby-coder.md
 create mode 100644 .claude/agents/ruby-reviewer.md
 create mode 100644 .claude/agents/screenshot-guardian.md
 create mode 100644 .claude/agents/tester.md
 create mode 100644 .claude/agents/tutorial-engineer.md
 create mode 100644 .claude/agents/ui-problem-diagnosis.md
 create mode 100644 .claude/agents/ux-browser-validator.md
 create mode 100644 .claude/commands/blog-pipeline.md
 create mode 100644 .claude/commands/cover-image.md
 create mode 100644 .claude/commands/research.md
 create mode 100644 .claude/rules/css-consolidation.md
 create mode 100644 .claude/rules/icp-target.md
 create mode 100644 .claude/rules/project-rules.md
 create mode 100644 .claude/rules/tdd-enforcement.md
 create mode 100644 .claude/rules/voice-guide.md
 create mode 100644 .claude/skills/blog-post-pipeline/SKILL.md
 create mode 100644 .claude/skills/content-voice-enforcement/SKILL.md
 create mode 100644 .claude/skills/cover-image-generation/SKILL.md
 create mode 100644 .claude/skills/css-bem-migration/SKILL.md
 create mode 100644 .claude/skills/visual-regression-testing/SKILL.md
 create mode 100644 .codex/agents/analyst.toml
 create mode 100644 .codex/agents/build-monitor.toml
 create mode 100644 .codex/agents/coder.toml
 create mode 100644 .codex/agents/content-creator.toml
 create mode 100644 .codex/agents/content-marketer.toml
 create mode 100644 .codex/agents/hugo-expert.toml
 create mode 100644 .codex/agents/planner.toml
 create mode 100644 .codex/agents/qa-browser-tester.toml
 create mode 100644 .codex/agents/researcher.toml
 create mode 100644 .codex/agents/reviewer.toml
 create mode 100644 .codex/agents/ruby-coder.toml
 create mode 100644 .codex/agents/ruby-reviewer.toml
 create mode 100644 .codex/agents/screenshot-guardian.toml
 create mode 100644 .codex/agents/seo-specialist.toml
 create mode 100644 .codex/agents/tester.toml
 create mode 100644 .codex/agents/tutorial-engineer.toml
 create mode 100644 .codex/agents/ui-problem-diagnosis.toml
 create mode 100644 .codex/agents/ux-browser-validator.toml
 create mode 100644 .codex/config.toml
 create mode 100644 .cursor/agents/analyst.md
 create mode 100644 .cursor/agents/build-monitor.md
 create mode 100644 .cursor/agents/coder.md
 create mode 100644 .cursor/agents/content-creator.md
 create mode 100644 .cursor/agents/content-marketer.md
 create mode 100644 .cursor/agents/hugo-expert.md
 create mode 100644 .cursor/agents/planner.md
 create mode 100644 .cursor/agents/qa-browser-tester.md
 create mode 100644 .cursor/agents/researcher.md
 create mode 100644 .cursor/agents/reviewer.md
 create mode 100644 .cursor/agents/ruby-coder.md
 create mode 100644 .cursor/agents/ruby-reviewer.md
 create mode 100644 .cursor/agents/screenshot-guardian.md
 create mode 100644 .cursor/agents/seo-specialist.md
 create mode 100644 .cursor/agents/tester.md
 create mode 100644 .cursor/agents/tutorial-engineer.md
 create mode 100644 .cursor/agents/ui-problem-diagnosis.md
 create mode 100644 .cursor/agents/ux-browser-validator.md
 create mode 100644 .cursor/commands/blog-pipeline.md
 create mode 100644 .cursor/commands/cover-image.md
 create mode 100644 .cursor/commands/research.md
 create mode 100644 .cursor/mcp.json
 create mode 100644 .cursor/rules/css-consolidation.mdc
 create mode 100644 .cursor/rules/icp-target.mdc
 create mode 100644 .cursor/rules/project-rules.mdc
 create mode 100644 .cursor/rules/tdd-enforcement.mdc
 create mode 100644 .cursor/rules/voice-guide.mdc
 create mode 100644 .gemini/commands/blog-pipeline.toml
 create mode 100644 .gemini/commands/cover-image.toml
 create mode 100644 .gemini/commands/research.toml
 create mode 100644 .gemini/settings.json
 create mode 100644 .github/agents/analyst.agent.md
 create mode 100644 .github/agents/build-monitor.agent.md
 create mode 100644 .github/agents/coder.agent.md
 create mode 100644 .github/agents/content-creator.agent.md
 create mode 100644 .github/agents/content-marketer.agent.md
 create mode 100644 .github/agents/hugo-expert.agent.md
 create mode 100644 .github/agents/planner.agent.md
 create mode 100644 .github/agents/qa-browser-tester.agent.md
 create mode 100644 .github/agents/researcher.agent.md
 create mode 100644 .github/agents/reviewer.agent.md
 create mode 100644 .github/agents/ruby-coder.agent.md
 create mode 100644 .github/agents/ruby-reviewer.agent.md
 create mode 100644 .github/agents/screenshot-guardian.agent.md
 create mode 100644 .github/agents/seo-specialist.agent.md
 create mode 100644 .github/agents/tester.agent.md
 create mode 100644 .github/agents/tutorial-engineer.agent.md
 create mode 100644 .github/agents/ui-problem-diagnosis.agent.md
 create mode 100644 .github/agents/ux-browser-validator.agent.md
 create mode 100644 .github/instructions/css-consolidation.instructions.md
 create mode 100644 .github/instructions/icp-target.instructions.md
 create mode 100644 .github/instructions/project-rules.instructions.md
 create mode 100644 .github/instructions/tdd-enforcement.instructions.md
 create mode 100644 .github/instructions/voice-guide.instructions.md
 create mode 100644 .github/prompts/blog-pipeline.prompt.md
 create mode 100644 .github/prompts/cover-image.prompt.md
 create mode 100644 .github/prompts/research.prompt.md
 create mode 100644 .kiro/settings/mcp.json
 create mode 100644 .kiro/skills/blog-post-pipeline/SKILL.md
 create mode 100644 .kiro/skills/content-voice-enforcement/SKILL.md
 create mode 100644 .kiro/skills/cover-image-generation/SKILL.md
 create mode 100644 .kiro/skills/css-bem-migration/SKILL.md
 create mode 100644 .kiro/skills/visual-regression-testing/SKILL.md
 create mode 100644 .kiro/steering/css-consolidation.md
 create mode 100644 .kiro/steering/icp-target.md
 create mode 100644 .kiro/steering/project-rules.md
 create mode 100644 .kiro/steering/tdd-enforcement.md
 create mode 100644 .kiro/steering/voice-guide.md
 create mode 100644 .opencode/agents/analyst.md
 create mode 100644 .opencode/agents/build-monitor.md
 create mode 100644 .opencode/agents/coder.md
 create mode 100644 .opencode/agents/content-creator.md
 create mode 100644 .opencode/agents/content-marketer.md
 create mode 100644 .opencode/agents/hugo-expert.md
 create mode 100644 .opencode/agents/planner.md
 create mode 100644 .opencode/agents/qa-browser-tester.md
 create mode 100644 .opencode/agents/researcher.md
 create mode 100644 .opencode/agents/reviewer.md
 create mode 100644 .opencode/agents/ruby-coder.md
 create mode 100644 .opencode/agents/ruby-reviewer.md
 create mode 100644 .opencode/agents/screenshot-guardian.md
 create mode 100644 .opencode/agents/seo-specialist.md
 create mode 100644 .opencode/agents/tester.md
 create mode 100644 .opencode/agents/tutorial-engineer.md
 create mode 100644 .opencode/agents/ui-problem-diagnosis.md
 create mode 100644 .opencode/agents/ux-browser-validator.md
 create mode 100644 .opencode/commands/blog-pipeline.md
 create mode 100644 .opencode/commands/cover-image.md
 create mode 100644 .opencode/commands/research.md
 create mode 100644 .windsurf/rules/css-consolidation.md
 create mode 100644 .windsurf/rules/icp-target.md
 create mode 100644 .windsurf/rules/project-rules.md
 create mode 100644 .windsurf/rules/tdd-enforcement.md
 create mode 100644 .windsurf/rules/voice-guide.md
 create mode 100644 .windsurf/workflows/blog-pipeline.md
 create mode 100644 .windsurf/workflows/cover-image.md
 create mode 100644 .windsurf/workflows/research.md
 create mode 100644 Makefile
 create mode 100644 apm.lock.yaml
 create mode 100644 apm.yml
 create mode 100644 compose.mcp.yml
 create mode 100644 docs/mcp-user-setup.md
 create mode 100644 opencode.json

diff --git a/.agents/skills/blog-post-pipeline/SKILL.md b/.agents/skills/blog-post-pipeline/SKILL.md
new file mode 100644
index 000000000..5f2324f46
--- /dev/null
+++ b/.agents/skills/blog-post-pipeline/SKILL.md
@@ -0,0 +1,41 @@
+# Blog Post Pipeline Skill
+#
+# Reusable 9-phase autonomous content pipeline for Hugo-based blogs.
+# This is a PRODUCER skill — published by jetthoughts/site for other repos.
+
+## Overview
+End-to-end pipeline from topic selection to published post with automated
+quality gates, multi-critic review, and cover image generation.
+
+## Phases
+1. Pick topic from content plan
+2. Read context (ICP, voice guide, brief)
+3. Research (5+ authoritative sources)
+4. Draft with voice enforcement + cadence quotas
+5. Review loop (3 parallel critic agents: founder, SEO/slop, editor)
+5c. Cold-eyes gate (9 explicit checks)
+6. Cover image (6-slot HTML → chrome-devtools render → lanczos downsample)
+7. Validate (Hugo build + Chrome DevTools)
+8. Update content plan
+9. Commit
+
+## Dependencies
+### MCP Servers
+- claude-context — semantic codebase search for existing posts
+- brave-search / searxng — web research
+- chrome-devtools — cover image rendering, validation
+
+### Other Skills
+- content-voice-enforcement — voice guide compliance
+- cover-image-generation — 6-slot layout rendering
+
+### Agents
+- content-creator — drafting
+- seo-specialist — SEO compliance
+- reviewer — quality gates
+
+## Trigger
+@.apm/prompts/blog-pipeline.prompt.md
+
+## Full Documentation
+docs/workflows/blog-pipeline.md
diff --git a/.agents/skills/content-voice-enforcement/SKILL.md b/.agents/skills/content-voice-enforcement/SKILL.md
new file mode 100644
index 000000000..df01df43f
--- /dev/null
+++ b/.agents/skills/content-voice-enforcement/SKILL.md
@@ -0,0 +1,37 @@
+# Content Voice Enforcement Skill
+#
+# Reusable voice enforcement for JetThoughts content — blog posts, LinkedIn,
+# and marketing copy targeting the ICP-E audience (non-technical founder
+# burned by a devshop).
+
+## Overview
+Enforces the JetThoughts voice guide, banned word list, structural patterns,
+cadence quotas, and quality tests on all content.
+
+## Voice Guide
+See .apm/instructions/voice-guide.instructions.md for the full compiled extract.
+Canonical source: docs/90-99-content-strategy/strategy-analysis/90.11-voice-guide.md
+
+## Two Voices
+- **Founder posts**: Rob Walling directness + Rand Fishkin vulnerability
+- **Tech posts**: Thoughtbot style (tension-first, opinionated, code as evidence)
+
+## Quality Tests (run on every paragraph)
+- **"Who" test**: who is doing what?
+- **"Show" test**: did I show the mechanic or just label it?
+- **"Coffee" test**: would a tired founder talk like this at a bar?
+- **"90/10" test**: ≥90% educational, ≤10% promotional
+- **"Trade-off" test**: acknowledged what the solution fails to do?
+- **"Practitioner" test**: replace generalizations with specific incidents
+
+## Slop Detector + Cold-Eyes Gate
+- Run slop-detector on all drafts
+- Run humanizer pass if flagged
+- Cold-eyes gate (9 checks) mandatory before publishing
+- See docs/workflows/blog-pipeline.md STEP 5a + STEP 5c
+
+## Full Documentation
+.apm/instructions/voice-guide.instructions.md
+.apm/instructions/icp-target.instructions.md
+docs/90-99-content-strategy/strategy-analysis/90.11-voice-guide.md
+docs/90-99-content-strategy/strategy-analysis/90.10-icp-primary-website-target.md
diff --git a/.agents/skills/cover-image-generation/SKILL.md b/.agents/skills/cover-image-generation/SKILL.md
new file mode 100644
index 000000000..a367207dc
--- /dev/null
+++ b/.agents/skills/cover-image-generation/SKILL.md
@@ -0,0 +1,32 @@
+# Cover Image Generation Skill
+#
+# Reusable skill for generating blog post cover images using the
+# JetVelocity 6-slot design system.
+
+## Overview
+Generates 2400×1260 cover images for blog posts using a 6-slot HTML layout
+rendered via chrome-devtools.
+
+## Design System
+- JetVelocity: obsidian dark, Ruby red (#cc342d), neon purple (#a855f7)
+- Font: Inter (headlines), JetBrains Mono (code/metrics)
+- 6-slot grid layout — canonical spec in .stitch/design.md
+
+## Workflow
+1. Read .stitch/design.md for the 6-slot layout specification
+2. Duplicate an existing HTML from .stitch/designs/
+3. Change ONLY: title, category, pill, 3 headline lines, 3 chip labels+values, status text
+4. Render: chrome-devtools new_page → emulate viewport=2400x1260x2 → wait 2s → take_screenshot
+5. Downsample: magick {screenshot} -filter Lanczos -resize 2400x1260 -sampling-factor 4:4:4 -strip -quality 95 content/blog/<slug>/cover.png
+6. Set frontmatter: cover_image: cover.png, cover_image_alt: descriptive alt text
+
+## Dependencies
+### MCP
+- chrome-devtools — rendering
+
+### System
+- ImageMagick (magick CLI) — downsampling
+
+## Full Documentation
+.stitch/design.md (canonical 6-slot spec)
+docs/workflows/cover-images.md
diff --git a/.agents/skills/css-bem-migration/SKILL.md b/.agents/skills/css-bem-migration/SKILL.md
new file mode 100644
index 000000000..9fcd1046f
--- /dev/null
+++ b/.agents/skills/css-bem-migration/SKILL.md
@@ -0,0 +1,49 @@
+# CSS BEM Migration Skill
+#
+# Reusable methodology for migrating from FL-Builder (Beaver Builder)
+# auto-generated classes to BEM component architecture.
+
+## Overview
+TDD-driven FL-node extraction: add BEM classes → verify visual fidelity →
+remove FL-nodes → retest. Micro-commits on every green test.
+
+## Pattern
+1. RED: Identify FL-node CSS dependencies
+2. GREEN: Add BEM classes to HTML, consolidate CSS
+3. REFACTOR: Remove duplicate FL-node CSS, cleanup
+4. TEST: `env PRECOMPILED_ASSETS=false bin/rake test:critical`
+5. COMMIT: Micro-commit on green tests
+
+## BEM Component Structure
+```html
+<!-- Before (FL-node) -->
+<div class="fl-col fl-node-yx43bujcaiqn hover-box">
+
+<!-- Transition (BEM + FL retention) -->
+<div class="fl-col fl-col-small hover-box c-feature-card">
+
+<!-- Final (Pure BEM) -->
+<div class="c-feature-card c-feature-card--variant">
+```
+
+## CSS Consolidation Pattern
+```css
+/* Before: 3 separate FL-node selectors */
+.fl-node-yx43bujcaiqn { width: 32%; margin-right: 30px; }
+.fl-node-ktz4ipj39vd6 { width: 36%; margin: 0 30px; }
+.fl-node-m39uvorzy5g8 { width: 32%; margin-left: 30px; }
+
+/* After: DRY BEM pattern */
+.c-feature-card { width: 32%; margin-right: 30px; }
+.c-feature-card--second { width: 36%; margin: 0 30px; }
+.c-feature-card--third { margin-right: 0; }
+```
+
+## Constraints
+- Never modify vendor files or dynamic templates
+- Preserve cascade order: Critical → Layout → Component → Theme → Footer
+- Run BOTH bin/test (macOS) AND bin/dtest (Linux Docker) before commit
+
+## Full Documentation
+docs/workflows/css-consolidation.md
+docs/projects/2509-css-migration/css-loading-order-analysis.md
diff --git a/.agents/skills/visual-regression-testing/SKILL.md b/.agents/skills/visual-regression-testing/SKILL.md
new file mode 100644
index 000000000..ff12576f9
--- /dev/null
+++ b/.agents/skills/visual-regression-testing/SKILL.md
@@ -0,0 +1,41 @@
+# Visual Regression Testing Skill
+#
+# Reusable visual regression testing protocol for Hugo static sites.
+
+## Overview
+Screenshot comparison-based visual testing using capybara-screenshot-diff.
+Blocks commits on any visual regression during refactoring.
+
+## Protocol
+1. Capture baseline screenshots BEFORE changes
+2. Make CSS/HTML changes
+3. Capture comparison screenshots AFTER changes
+4. Compare with tolerance:
+   - 0.0 for refactoring (zero visual changes)
+   - ≤0.03 for new features only
+5. Block commit on any difference > tolerance
+
+## Test Pattern
+```ruby
+def test_hero_section_visual_consistency
+  visit "/services/fractional-cto/"
+  screenshot_path = take_screenshot("fractional_cto_hero")
+  assert_visual_match(screenshot_path, "baseline/fractional_cto_hero.png",
+                      tolerance: 0.03)
+end
+```
+
+## Prohibited Patterns (ZERO TOLERANCE)
+- CSS class existence testing
+- HTML structure counting
+- Framework-generated selector usage
+- Method existence testing
+
+## Commands
+- `bin/test` — visual regression (macOS)
+- `bin/dtest` — visual regression (Linux/Docker, CI parity)
+- BOTH must pass before commit for HTML/CSS changes
+
+## Full Documentation
+.apm/instructions/tdd-enforcement.instructions.md
+docs/60.03-tdd-quality-enforcement.md
diff --git a/.apm/agents/build-monitor.agent.md b/.apm/agents/build-monitor.agent.md
new file mode 100644
index 000000000..2bb9c0910
--- /dev/null
+++ b/.apm/agents/build-monitor.agent.md
@@ -0,0 +1,18 @@
+# build-monitor.agent.md — Thin Wrapper
+#
+# Full agent definitions: .claude/agents/build-monitor.md
+
+## Role
+Build monitoring agent. Validates Hugo builds, CI/CD pipelines, and performance.
+
+## Instructions
+- @.apm/instructions/project-rules.instructions.md
+
+## Gates
+- `bin/hugo-build` must pass (zero errors)
+- Chrome DevTools: zero console errors, zero 404s
+- `bin/rake test:critical` green on every commit
+
+## CI/CD
+- GitHub Actions → GitHub Pages
+- Workflows: .github/workflows/publish.yml, test.yml, sync-and-publish.yml
diff --git a/.apm/agents/content/content-creator.agent.md b/.apm/agents/content/content-creator.agent.md
new file mode 100644
index 000000000..e6c244ca8
--- /dev/null
+++ b/.apm/agents/content/content-creator.agent.md
@@ -0,0 +1,20 @@
+# content-creator.agent.md — Thin Wrapper
+#
+# Canonical instructions: .apm/instructions/voice-guide.instructions.md
+# Full agent definitions: .claude/agents/content-creator.md
+
+## Role
+Content creation agent. Writes blog posts, LinkedIn posts, and marketing copy.
+
+## Instructions
+- @.apm/instructions/voice-guide.instructions.md
+- @.apm/instructions/icp-target.instructions.md
+- @docs/workflows/blog-pipeline.md
+- @docs/workflows/linkedin-post-pipeline.md
+
+## Constraints
+- Zero generic AI language — run slop-detector + humanizer pass
+- Cold-eyes gate (9 checks) mandatory before publishing
+- Every sentence needs a person doing something
+- Banned words: unlock, harness, leverage, dive in, game-changer, etc.
+- Use `-` not `—` for all dashes
diff --git a/.apm/agents/content/content-marketer.agent.md b/.apm/agents/content/content-marketer.agent.md
new file mode 100644
index 000000000..053ecd0a7
--- /dev/null
+++ b/.apm/agents/content/content-marketer.agent.md
@@ -0,0 +1,10 @@
+# content-marketer.agent.md — Thin Wrapper
+#
+# Full agent definitions: .claude/agents/content/content-marketer.md
+
+## Role
+Content marketing agent. Plans content calendars, researches topics, audits SEO performance.
+
+## Instructions
+- @.apm/instructions/icp-target.instructions.md
+- @docs/projects/2510-seo-content-strategy/20-29-strategy/20.07-content-plan-icp-e-q2-2026.md
diff --git a/.apm/agents/content/seo-specialist.agent.md b/.apm/agents/content/seo-specialist.agent.md
new file mode 100644
index 000000000..328a80718
--- /dev/null
+++ b/.apm/agents/content/seo-specialist.agent.md
@@ -0,0 +1,21 @@
+# seo-specialist.agent.md — Thin Wrapper
+#
+# Full agent definitions: .claude/agents/seo-specialist.md
+
+## Role
+SEO optimization agent. Handles keyword research, meta tags, schema, and content optimization.
+
+## Instructions
+- @docs/90-99-content-strategy/seo-optimization/
+- @docs/90-99-content-strategy/strategy-analysis/
+
+## Key Files
+- `themes/beaver/layouts/partials/seo/enhanced-meta-tags.html` — SEO/og:image template
+- `themes/beaver/layouts/partials/seo/article-schema.html` — JSON-LD schema
+- `config/_default/hugo.toml` — Hugo config (permalinks, sitemap)
+
+## Constraints
+- Title ≤45 chars, description 150-160 chars
+- Primary keyword in first 100 words
+- 4+ internal links, 5+ external citations
+- No keyword stuffing
diff --git a/.apm/agents/content/tutorial-engineer.agent.md b/.apm/agents/content/tutorial-engineer.agent.md
new file mode 100644
index 000000000..14374cb5e
--- /dev/null
+++ b/.apm/agents/content/tutorial-engineer.agent.md
@@ -0,0 +1,10 @@
+# tutorial-engineer.agent.md — Thin Wrapper
+#
+# Full agent definitions: .claude/agents/content/tutorial-engineer.md
+
+## Role
+Tutorial engineer. Creates step-by-step how-to content, validates code examples, and ensures copy-paste-ability.
+
+## Instructions
+- @.apm/instructions/voice-guide.instructions.md
+- @.apm/instructions/icp-target.instructions.md
diff --git a/.apm/agents/core/analyst.agent.md b/.apm/agents/core/analyst.agent.md
new file mode 100644
index 000000000..dfa4ca692
--- /dev/null
+++ b/.apm/agents/core/analyst.agent.md
@@ -0,0 +1,15 @@
+# analyst.agent.md — Thin Wrapper
+#
+# Full agent definitions: .claude/agents/core/analyst.md
+
+## Role
+Analysis agent. Performs research, root-cause analysis, and pattern discovery.
+
+## Instructions
+- @.apm/instructions/project-rules.instructions.md
+- @.apm/prompts/research.prompt.md
+
+## Research Protocol (MANDATORY)
+1. claude-context search FIRST
+2. Verify with rg/ls
+3. Read docs/workflows/BASE_HANDBOOK.md
diff --git a/.apm/agents/core/coder.agent.md b/.apm/agents/core/coder.agent.md
new file mode 100644
index 000000000..7c8b2615e
--- /dev/null
+++ b/.apm/agents/core/coder.agent.md
@@ -0,0 +1,26 @@
+# coder.agent.md — Thin Wrapper
+#
+# Canonical instructions: .apm/instructions/project-rules.instructions.md
+# Full agent definitions: .claude/agents/core/coder.md
+# 
+# APM compiles this agent for all target tools. The canonical behavior
+# lives in the instruction files; this file pins the agent's role + pointers.
+
+## Role
+Primary implementation agent. Writes code following TDD cycle (RED → GREEN → REFACTOR).
+
+## Instructions
+- @.apm/instructions/project-rules.instructions.md
+- @.apm/instructions/tdd-enforcement.instructions.md
+
+## For CSS/HTML work
+- @.apm/instructions/css-consolidation.instructions.md
+
+## For Ruby work
+- @docs/workflows/ruby.md
+
+## Constraints
+- Behavior-focused tests ONLY — reject structural/CSS class tests
+- Run `bin/rake test:critical` after every micro-change
+- Never use `npm` — use `bun`
+- Never create ad hoc `*.sh` test scripts
diff --git a/.apm/agents/core/planner.agent.md b/.apm/agents/core/planner.agent.md
new file mode 100644
index 000000000..837cedb13
--- /dev/null
+++ b/.apm/agents/core/planner.agent.md
@@ -0,0 +1,9 @@
+# planner.agent.md — Thin Wrapper
+#
+# Full agent definitions: .claude/agents/core/planner.md
+
+## Role
+Planning agent. Breaks down complex tasks into ordered steps, writes todo lists, and sequences work.
+
+## Instructions
+- @.apm/instructions/project-rules.instructions.md
diff --git a/.apm/agents/core/researcher.agent.md b/.apm/agents/core/researcher.agent.md
new file mode 100644
index 000000000..5d953465d
--- /dev/null
+++ b/.apm/agents/core/researcher.agent.md
@@ -0,0 +1,10 @@
+# researcher.agent.md — Thin Wrapper
+#
+# Full agent definitions: .claude/agents/core/researcher.md
+
+## Role
+Research agent. Discovers existing patterns, framework docs, and best practices before implementation.
+
+## Instructions
+- @.apm/instructions/project-rules.instructions.md
+- @.apm/prompts/research.prompt.md
diff --git a/.apm/agents/core/reviewer.agent.md b/.apm/agents/core/reviewer.agent.md
new file mode 100644
index 000000000..989624346
--- /dev/null
+++ b/.apm/agents/core/reviewer.agent.md
@@ -0,0 +1,17 @@
+# reviewer.agent.md — Thin Wrapper
+#
+# Canonical instructions: .apm/instructions/project-rules.instructions.md
+# Full agent definitions: .claude/agents/core/reviewer.md
+
+## Role
+Code review agent. Validates correctness, adherence to project rules, and TDD compliance.
+
+## Instructions
+- @.apm/instructions/project-rules.instructions.md
+- @.apm/instructions/tdd-enforcement.instructions.md
+
+## Gates
+- Zero structural test patterns (CSS class testing, HTML counting)
+- Visual regression tolerance verified
+- Four-eyes principle enforced for all critical changes
+- Feature branch + PR pattern for HTML/CSS/template changes
diff --git a/.apm/agents/core/screenshot-guardian.agent.md b/.apm/agents/core/screenshot-guardian.agent.md
new file mode 100644
index 000000000..f160b3b75
--- /dev/null
+++ b/.apm/agents/core/screenshot-guardian.agent.md
@@ -0,0 +1,15 @@
+# screenshot-guardian.agent.md — Thin Wrapper
+#
+# Full agent definitions: .claude/agents/core/screenshot-guardian.md
+
+## Role
+Visual regression guardian. Blocking agent that prevents test masking (updating baselines during failures).
+
+## Instructions
+- @.apm/instructions/tdd-enforcement.instructions.md
+- @.claude/hooks/pre-commit-screenshot-validation.md
+
+## Gates
+- ZERO tolerance: screenshot updates during visual test failures
+- Tolerance: 0.0 for refactors, ≤0.03 for new features
+- Both bin/test AND bin/dtest must pass before commit
diff --git a/.apm/agents/core/tester.agent.md b/.apm/agents/core/tester.agent.md
new file mode 100644
index 000000000..8bc5f213b
--- /dev/null
+++ b/.apm/agents/core/tester.agent.md
@@ -0,0 +1,25 @@
+# tester.agent.md — Thin Wrapper
+#
+# Canonical instructions: .apm/instructions/project-rules.instructions.md
+# Full agent definitions: .claude/agents/core/tester.md
+
+## Role
+Test agent. Writes and runs tests following TDD methodology.
+
+## Instructions
+- @.apm/instructions/project-rules.instructions.md
+- @.apm/instructions/tdd-enforcement.instructions.md
+
+## Framework
+- Minitest (test/system/, test/unit/)
+- capybara-screenshot-diff for visual regression
+
+## Commands
+- `bin/rake test:critical` — critical tests
+- `bin/test` — visual regression (macOS)
+- `bin/dtest` — visual regression (Linux/Docker)
+
+## Constraints
+- Behavior-focused tests ONLY
+- Assert shape, not config values
+- Run both bin/test AND bin/dtest before commit for HTML/CSS changes
diff --git a/.apm/agents/hugo-expert.agent.md b/.apm/agents/hugo-expert.agent.md
new file mode 100644
index 000000000..24fe61525
--- /dev/null
+++ b/.apm/agents/hugo-expert.agent.md
@@ -0,0 +1,25 @@
+# hugo-expert.agent.md — Thin Wrapper
+#
+# Full agent definitions: .claude/agents/hugo-expert.md
+
+## Role
+Hugo static site generator expert. Handles templates, shortcodes, content structure, and build optimization.
+
+## Instructions
+- @.apm/instructions/project-rules.instructions.md
+- @docs/workflows/css-consolidation.md
+
+## Key Files
+- `config/_default/hugo.toml` — Hugo configuration
+- `themes/beaver/layouts/` — Templates
+- `themes/beaver/assets/css/` — CSS (PostCSS pipeline)
+- `content/` — Content files (Markdown + Hugo frontmatter)
+
+## Commands
+- `bin/hugo-build` — Build + validate
+- `hugo server -D` — Dev server
+
+## Constraints
+- Never modify vendor CSS files
+- Preserve cascade order: Critical → Layout → Component → Theme → Footer
+- BEM methodology for all new CSS
diff --git a/.apm/agents/ruby/ruby-coder.agent.md b/.apm/agents/ruby/ruby-coder.agent.md
new file mode 100644
index 000000000..510d186df
--- /dev/null
+++ b/.apm/agents/ruby/ruby-coder.agent.md
@@ -0,0 +1,15 @@
+# ruby-coder.agent.md — Thin Wrapper
+#
+# Full agent definitions: .claude/agents/ruby/ruby-coder.md
+
+## Role
+Ruby implementation agent. Writes Ruby code (tests, sync scripts, lib/) following TDD.
+
+## Instructions
+- @.apm/instructions/project-rules.instructions.md
+- @.apm/instructions/tdd-enforcement.instructions.md
+- @docs/workflows/ruby.md
+
+## Commands
+- `bin/rake test:critical`
+- `bundle exec standardrb --fix`
diff --git a/.apm/agents/ruby/ruby-reviewer.agent.md b/.apm/agents/ruby/ruby-reviewer.agent.md
new file mode 100644
index 000000000..bdfcf27b8
--- /dev/null
+++ b/.apm/agents/ruby/ruby-reviewer.agent.md
@@ -0,0 +1,14 @@
+# ruby-reviewer.agent.md — Thin Wrapper
+#
+# Full agent definitions: .claude/agents/ruby/ruby-reviewer.md
+
+## Role
+Ruby code reviewer. Validates Ruby code quality, TDD compliance, and Minitest patterns.
+
+## Instructions
+- @.apm/instructions/tdd-enforcement.instructions.md
+
+## Constraints
+- Behavior-focused tests ONLY — reject structural testing
+- StandardRB compliance
+- No ad hoc test scripts
diff --git a/.apm/agents/validation/qa-browser-tester.agent.md b/.apm/agents/validation/qa-browser-tester.agent.md
new file mode 100644
index 000000000..3a18ca639
--- /dev/null
+++ b/.apm/agents/validation/qa-browser-tester.agent.md
@@ -0,0 +1,16 @@
+# qa-browser-tester.agent.md — Thin Wrapper
+#
+# Full agent definitions: .claude/agents/validation/qa-browser-tester.md
+
+## Role
+QA browser tester. Runs Chrome DevTools validation after HTML/CSS/JS changes.
+
+## Instructions
+- @.apm/instructions/project-rules.instructions.md
+- @.apm/instructions/tdd-enforcement.instructions.md
+
+## Gates
+- Console: ZERO JS errors
+- Network: ZERO 404s for assets
+- Capture desktop + mobile screenshots
+- All gates pass → commit
diff --git a/.apm/agents/validation/ui-problem-diagnosis.agent.md b/.apm/agents/validation/ui-problem-diagnosis.agent.md
new file mode 100644
index 000000000..9f88e69ee
--- /dev/null
+++ b/.apm/agents/validation/ui-problem-diagnosis.agent.md
@@ -0,0 +1,10 @@
+# ui-problem-diagnosis.agent.md — Thin Wrapper
+#
+# Full agent definitions: .claude/agents/validation/ui-problem-diagnosis-specialist.md
+
+## Role
+UI problem diagnosis specialist. Investigates visual regressions, cross-browser issues, and CSS bugs.
+
+## Instructions
+- @.apm/instructions/css-consolidation.instructions.md
+- @docs/projects/2509-css-migration/css-loading-order-analysis.md
diff --git a/.apm/agents/validation/ux-browser-validator.agent.md b/.apm/agents/validation/ux-browser-validator.agent.md
new file mode 100644
index 000000000..118ea4ffa
--- /dev/null
+++ b/.apm/agents/validation/ux-browser-validator.agent.md
@@ -0,0 +1,14 @@
+# ux-browser-validator.agent.md — Thin Wrapper
+#
+# Full agent definitions: .claude/agents/validation/ux-browser-validator.md
+
+## Role
+UX browser validator. Validates UX interactions, accessibility, and cross-browser behavior.
+
+## Instructions
+- @.apm/instructions/project-rules.instructions.md
+
+## Gates
+- WCAG 2.1 AA accessibility compliance
+- Semantic HTML elements (h1, button, main, nav)
+- Responsive design: desktop + mobile verification
diff --git a/.apm/instructions/css-consolidation.instructions.md b/.apm/instructions/css-consolidation.instructions.md
new file mode 100644
index 000000000..0f1faae2d
--- /dev/null
+++ b/.apm/instructions/css-consolidation.instructions.md
@@ -0,0 +1,30 @@
+# CSS Consolidation Rules — JetThoughts Site
+#
+# MANDATORY for all CSS/HTML template changes.
+# Full docs: docs/workflows/css-consolidation.md
+# Project: docs/projects/2509-css-migration/css-loading-order-analysis.md
+
+## Architecture
+- **Methodology**: BEM (Block__Element--Modifier)
+- **Pipeline**: PostCSS (import, mixins, nested, autoprefixer, cssnano, purgecss)
+- **Cascade order**: Critical → Layout → Component → Theme → Footer
+
+## Rules
+- Never modify vendor files (`css/vendors/*.min.css`) or dynamic templates (`css/dynamic-*.css`)
+- Preserve cascade order in all changes
+- Extract whole rule sets; keep page-specific `.fl-node-{hash}` in original files
+- Run `bin/rake test:critical` after each extraction
+
+## FL-Node Extraction (Beaver Builder → BEM)
+- **Phase**: TDD RED-GREEN-REFACTOR cycle
+- **Validation**: `env PRECOMPILED_ASSETS=false bin/rake test:critical` after each change
+- **Pattern**: Add BEM classes first → verify visual fidelity → remove FL-nodes → retest
+- **Commit**: micro-commit on green tests only
+
+## Visual Regression
+- Tolerance: 0.0 for refactoring (zero visual changes)
+- Tolerance: ≤0.03 for new features only
+- Run BOTH `bin/test` and `bin/dtest` before commit
+- Visual failures are commit blockers, not warnings
+
+@see docs/workflows/css-consolidation.md
diff --git a/.apm/instructions/icp-target.instructions.md b/.apm/instructions/icp-target.instructions.md
new file mode 100644
index 000000000..e9457d522
--- /dev/null
+++ b/.apm/instructions/icp-target.instructions.md
@@ -0,0 +1,46 @@
+# ICP Target — JetThoughts Primary Website Audience
+#
+# CANONICAL SOURCE: docs/90-99-content-strategy/strategy-analysis/90.10-icp-primary-website-target.md
+# This file is a COMPILED EXTRACT for APM — the full doc is authoritative.
+# Update the canonical source first, then sync this extract.
+#
+# MANDATORY reading before content, design, marketing, or blog work.
+# Compiled by APM into every AI tool's context.
+
+## ICP-E: Non-Technical Founder Burned by a Devshop
+
+### Who They Are
+- 35-50 years old
+- Non-technical or semi-technical
+- Raised funding ($500K–$2M seed)
+- Hired a dev shop or freelancers
+- Missed deadlines, buggy product, mounting technical debt
+- Starting to question: "Is it me? Is it them? How do I know?"
+
+### Their Pain
+- Can't evaluate code quality themselves
+- Dev shop says "almost done" for months
+- Every new feature breaks something else
+- No tests, no documentation, no deploy pipeline
+- Starting to lose investor confidence
+
+### What They Need
+- Someone who speaks plain English, not jargon
+- Proof that the problem is real and solvable (specific stories, not vague promises)
+- A clear path out: audit → stabilize → build right
+- Trust signals: named clients, specific results, honest trade-offs
+
+### What Resonates
+- Stories of other founders in the same situation
+- Specific numbers: "$40K wasted," "11-minute outage," "84 support tickets"
+- Honest acknowledgment that rescue is hard and expensive
+- The idea that good engineering is a competitive advantage
+- "You're not crazy — this IS what bad engineering looks like"
+
+### What Alienates
+- jargon-dense technical explanations before establishing trust
+- "We're the best" claims without evidence
+- Generic consulting language ("partner with you," "drive outcomes")
+- Ignoring the emotional reality (shame, frustration, fear of looking stupid)
+
+@see docs/90-99-content-strategy/strategy-analysis/90.10-icp-primary-website-target.md
diff --git a/.apm/instructions/project-rules.instructions.md b/.apm/instructions/project-rules.instructions.md
new file mode 100644
index 000000000..deb7313e4
--- /dev/null
+++ b/.apm/instructions/project-rules.instructions.md
@@ -0,0 +1,79 @@
+# JetThoughts Site — Project Rules
+#
+# This is the SINGLE SOURCE for AI agent instructions across all tools.
+# APM compiles this to each tool's expected path (CLAUDE.md, .github/
+# copilot-instructions.md, etc.). Edit here, not in tool-specific files.
+#
+# CANONICAL SOURCE: THIS FILE. The existing CLAUDE.md is a Claude-specific
+# superset; over time CLI tool instructions converge here. Voice, ICP, TDD,
+# and CSS rules are split into separate instruction files for modularity.
+#
+# Full workflow docs: docs/workflows/
+# Full content docs: docs/90-99-content-strategy/
+
+## Stack
+- **Site**: Hugo 0.147.9 Extended static site generator
+- **Theme**: Custom "Beaver" theme (BEM + PostCSS)
+- **Tests**: Ruby 3.4, Minitest, Capybara, Selenium WebDriver
+- **CSS**: PostCSS pipeline (autoprefixer, cssnano, purgecss)
+- **Package managers**: Bun (JS, NOT npm), Bundler (Ruby)
+- **CI/CD**: GitHub Actions → GitHub Pages
+
+## Build & Test Commands
+- Build: `bin/hugo-build`
+- Test (critical): `bin/rake test:critical`
+- Test (visual/macOS): `bin/test`
+- Test (visual/Linux Docker): `bin/dtest`
+- Lint Ruby: `bundle exec standardrb --fix`
+
+## Research Protocol (MANDATORY — before any change)
+1. Search codebase with `claude-context` MCP FIRST
+2. Verify exact filenames with `rg`/`ls` as fallback
+3. Read `docs/workflows/BASE_HANDBOOK.md` at session start
+4. Read `docs/workflows/flow-router.md` to route to correct workflow
+5. For blog content: find existing posts via `docs/blog-post-index.md`
+6. Never guess slugs — verify with `ls content/blog/<slug>/index.md`
+
+## Non-Negotiables
+- **Zero generic AI language** in content — all AI-sounding phrases rejected
+- **Zero Hugo build breaks** — `bin/hugo-build` must pass for content changes
+- **Zero unsupported claims** — all assertions must have citations
+- **No ad hoc test scripts** — use `bin/rake test:critical` and `bin/test`
+- **No Python for analysis** — prefer `rg`, `sed`, `awk`, shell tools
+- **No duplicate throwaway files** — no `*_refactored.*`, `*_new.*`, `*_backup.*`
+- **New docs only under `docs/workflows/`** — else edit existing files
+- **Never commit coordinator/verification reports** — write to `/tmp/` instead
+
+## Content Work (Blog Posts, LinkedIn, Pages)
+- **BEFORE writing**: Read voice guide + ICP target (see instructions/voice-guide, instructions/icp-target)
+- **Blog posts**: Follow `docs/workflows/blog-pipeline.md` end-to-end
+- **LinkedIn**: Follow `docs/workflows/linkedin-post-pipeline.md`
+- **Cover images**: Follow `docs/workflows/cover-images.md` + `.stitch/design.md`
+- Run slop-detector + humanizer pass on all drafts
+- Cold-eyes gate (9 checks) mandatory before publishing
+
+## CSS / HTML Changes
+- **BEM methodology** mandatory for all new CSS
+- **FL-node extraction**: TDD cycle — RED → GREEN → REFACTOR → TEST
+- **Visual regression tolerance**: 0.0 for refactors, ≤0.03 for new features
+- **MANDATORY**: Run BOTH `bin/test` (macOS) AND `bin/dtest` (Linux Docker) before commit
+- **Feature branches**: HTML/CSS/template changes → feature branch → one PR per sprint
+- Zero tolerance: CSS class existence testing, framework selector usage in tests
+
+## TDD & Testing
+- **Cycle**: RED → GREEN → REFACTOR
+- **Framework**: Minitest (test/system/, test/unit/)
+- **Behavior-focused tests ONLY** — reject implementation/existence/config tests
+- **Avoid fragile config assertions** — assert shape, not tunable values
+- **Visual regression**: screenshot comparison, 0.0 tolerance for refactoring
+- Run `bin/rake test:critical` after every micro-change (<10 lines)
+- Chrome DevTools validation after HTML/CSS/JS changes: zero console errors, zero 404s
+
+## Behavioral Constraints
+- **User says "code is bad" or "over-engineered"**: HALT, 5-Why analysis, fix config, proceed
+- **Reflection triggers**: zero-defect tracking, test masking detection
+- **Four-eyes principle**: paired agents for all critical changes
+- **Direct-to-master only for**: content-only blog edits, commit fixes, docs, CLAUDE.md policy updates
+
+@see docs/workflows/ for full workflow documentation
+@see docs/90-99-content-strategy/ for content strategy
diff --git a/.apm/instructions/tdd-enforcement.instructions.md b/.apm/instructions/tdd-enforcement.instructions.md
new file mode 100644
index 000000000..a532cfa19
--- /dev/null
+++ b/.apm/instructions/tdd-enforcement.instructions.md
@@ -0,0 +1,62 @@
+# TDD Enforcement — JetThoughts Site
+#
+# MANDATORY for all code changes. Extracted from CLAUDE.md TDD section.
+# Full docs: docs/60.03-tdd-quality-enforcement.md, docs/60.04-four-eyes-principle.md
+
+## TDD Cycle
+1. **RED**: Write a failing test that validates BEHAVIOR (not implementation)
+2. **GREEN**: Minimal code to make the test pass
+3. **REFACTOR**: Clean up while tests stay green
+4. **COMMIT**: Micro-commit on green tests
+
+## Test Quality Rules (ZERO TOLERANCE)
+
+### Prohibited Test Patterns — Agent MUST Reject
+- ❌ CSS class existence testing (`.fl-node-*`, `.has_css?(".px-4")`)
+- ❌ HTML structure counting (`page.all("div").count`)
+- ❌ Framework-generated selector usage (FL-Builder, Tailwind utility classes)
+- ❌ Method existence testing (`assert SomeClass.method_defined?`)
+- ❌ Config value assertions (hardcoded `q=90`, exact file sizes, specific dimensions)
+- ❌ Implementation detail testing
+
+### Required Test Patterns
+- ✅ Visual regression: `take_screenshot` → `assert_visual_match` with tolerance
+- ✅ User-visible content: `assert page.has_content?("text")`
+- ✅ Semantic elements: `assert page.has_css?("h1", text: "...")`
+- ✅ User interaction: `click_button` → `assert_current_path`
+- ✅ Shape assertions: `q=\d+`, `src contains wsrv.nl`, not `q=90`, not exact URLs
+
+### Avoid Fragile Config Assertions
+Don't hardcode tunable values. Assert the *shape*, not the configuration:
+- Bad: `assert_match /q=90/, src`
+- Good: `assert_match /q=\d+/, src`
+- Bad: `assert_match /w=360/, src`
+- Good: `assert has_css?("picture source")`
+
+## Visual Regression Gates (BLOCKING for HTML/CSS changes)
+- **Tolerance**: 0.0 for refactoring, ≤0.03 for new features
+- **Protocol**: Capture baseline BEFORE changes → compare AFTER → block on any diff >0%
+- **MANDATORY**: Run BOTH `bin/test` (macOS) AND `bin/dtest` (Linux Docker) before commit
+- Visual regression failures are COMMIT BLOCKERS, not warnings
+
+## Test Runner
+- Framework: Minitest
+- Critical tests: `bin/rake test:critical` (after every micro-change <10 lines)
+- Visual macOS: `bin/test`
+- Visual Linux/Docker: `bin/dtest`
+- NEVER create ad hoc `*.sh` test scripts
+
+## Chrome DevTools Validation (MANDATORY after HTML/CSS/JS)
+1. Start Hugo dev server, open page in Chrome DevTools
+2. Console: ZERO JS errors
+3. Network: ZERO 404s
+4. Capture desktop + mobile screenshots
+5. All gates pass → commit. Any gate fails → fix, repeat
+
+## Feature Branch + PR (HTML/CSS/template changes)
+1. `git checkout -b <sprint-name>`
+2. Multiple related commits on branch
+3. Both test gates green on each commit
+4. `git push -u origin <branch>`
+5. ONE PR per sprint via `gh pr create`
+6. Direct-to-master only for: content-only blog edits, commit fixes, docs, CLAUDE.md updates
diff --git a/.apm/instructions/voice-guide.instructions.md b/.apm/instructions/voice-guide.instructions.md
new file mode 100644
index 000000000..97aeec148
--- /dev/null
+++ b/.apm/instructions/voice-guide.instructions.md
@@ -0,0 +1,68 @@
+# Voice Guide — JetThoughts Content
+#
+# CANONICAL SOURCE: docs/90-99-content-strategy/strategy-analysis/90.11-voice-guide.md
+# This file is a COMPILED EXTRACT for APM — the full doc is authoritative.
+# Update the canonical source first, then sync this extract.
+#
+# MANDATORY reading before writing ANY blog post, LinkedIn post, or marketing copy.
+# Compiled by APM into every AI tool's context.
+
+## Two Voices
+
+### Founder Posts (Paul Keen / JT leadership)
+- **Rob Walling directness** + **Rand Fishkin vulnerability**
+- Plain English, empathy, specific numbers
+- Talk to a non-technical founder burned by a devshop
+- "Been there" energy — practitioner not vendor
+- Trade-offs acknowledged honestly
+
+### Tech Posts (engineering team)
+- **Thoughtbot style**: open with tension, not features
+- Own your opinions — no "it depends" hedging
+- Code as narrative evidence, not decoration
+- Honest trade-offs: what does this solution fail to do?
+
+## Banned Words
+unlock, harness, leverage, dive in, game-changer, journey, seamless, robust,
+supercharge, revolutionize, embark, delve, discover, blast radius, suspect list
+
+## Banned Structural Patterns
+- Rule of three (parallel triads)
+- Signposting ("This post is about...", "Here's what you need to know")
+- Bold inline-header lists
+- Triple rhetorical questions
+- Slogany parallel closings ("X is A. Y is B.")
+- Therapist voice ("That's the hardest part")
+- Copula avoidance ("serves as" → use "is")
+- Impersonal fragment stacking — every sentence needs a human subject
+- Sustained staccato (3+ consecutive short fragments or uniform-length sentences)
+- Command structure repetition (3+ paragraphs starting with imperatives)
+- Apologetic caveats ("every project is different")
+- Telling instead of showing — replace adjectives with concrete mechanics
+- Fake authority generalizations ("We've seen this on every codebase")
+- Timeline fabrication — verify technology release dates before using "last year"
+- Use `-` not `—` for all dashes
+
+## Quality Tests (run on every paragraph)
+- **"Who" test**: who is doing what? If subject is a thing, rewrite with a person
+- **"Show" test**: did I show the mechanic or just label it?
+- **"Coffee" test**: would a tired founder talk like this at a bar?
+- **"90/10" test**: is ≥90% educational? Is promotion ≤10%?
+- **"Trade-off" test**: did I acknowledge what the solution fails to do?
+- **"Practitioner" test**: replace generalized scenarios with specific incidents
+
+## Mannerism Noun Cap
+discipline, cycle, pattern, canonical, workflow, framework, cadence, rhythm,
+contract, gate, gatekeeper, layer, round, review, pass
+→ Max 5 occurrences total per post. Vary to the concrete thing happening.
+
+## Cadence Quotas (per H2 section)
+- At least one sentence <9 words AND one >22 words
+- No 3 consecutive sentences within 3 words of each other
+- One single-sentence paragraph per H2
+- Max 3 sentences / 70 words per paragraph
+- No two sentences in same paragraph starting with same first word
+- "The..." paragraph openers: <20% of total paragraphs
+- "We..." paragraph openers: <15% of total paragraphs
+
+@see docs/90-99-content-strategy/strategy-analysis/90.11-voice-guide.md
diff --git a/.apm/prompts/blog-pipeline.prompt.md b/.apm/prompts/blog-pipeline.prompt.md
new file mode 100644
index 000000000..9ad0669b1
--- /dev/null
+++ b/.apm/prompts/blog-pipeline.prompt.md
@@ -0,0 +1,75 @@
+# Blog Post Pipeline — Trigger Prompt
+#
+# Copy this prompt to write the next blog post from the content plan.
+# Full pipeline: docs/workflows/blog-pipeline.md
+#
+# This is a reusable PROMPT — APM compiles it to slash-commands for
+# each AI tool (e.g., /blog-post in Claude, #blog-post in Copilot).
+
+Write the next blog post following the full pipeline.
+
+## STEP 1 — Pick Topic
+Read the content plan: docs/projects/2510-seo-content-strategy/20-29-strategy/20.07-content-plan-icp-e-q2-2026.md
+Find the next 🔲 Planned topic. Confirm slug, primary keyword, target publish date.
+
+## STEP 2 — Read Context (mandatory, do not skip)
+- ICP: docs/90-99-content-strategy/strategy-analysis/90.10-icp-primary-website-target.md
+- Voice guide: docs/90-99-content-strategy/strategy-analysis/90.11-voice-guide.md
+- Content plan brief: read the topic's brief section
+- Design spec: .stitch/design.md
+
+## STEP 3 — Research
+- Web search for 5+ authoritative sources
+- Check existing posts for duplication: ls content/blog/ | grep -iE "<key-terms>"
+- Verify every statistic has a citable source
+
+## STEP 3b — Find Internal Posts to Reference
+- Use claude-context MCP search FIRST
+- For tag/slug lookups: docs/blog-post-index.md
+- Verify each slug: ls content/blog/<slug>/index.md
+- Collect at least 4 verified slugs before drafting
+
+## STEP 4 — Draft
+Create content/blog/<slug>/index.md with:
+- Frontmatter: title ≤45 chars, description 150-160 chars, date, slug, keywords, tags
+- Voice: plain English, empathy first, founder-to-founder tone
+- Structure: punchy hook → keyword in first 100 words → numbered H2s → "When NOT to" → JT proof → CTA → related reading → further reading
+- 4+ internal links in body, 5+ external citations
+- Paragraphs ≤3 sentences
+- Cut test: "what does the reader DO differently?" — delete if answer is "nothing new"
+
+## STEP 5 — Review Loop
+Spawn 3 parallel critic agents:
+1. Founder persona (Alex, burned CEO) — ≥8/10
+2. SEO/slop detector — ≥8/10 humanization
+3. Senior copy editor — "publish as-is" or "minor edits"
+
+## STEP 5c — Cold-Eyes Gate (mandatory)
+Spawn fresh-context reviewer with 9 explicit checks:
+1. Frontmatter vs body consistency
+2. Fluffy AI narration detection
+3. Fake authority stamps
+4. Timeline fabrication
+5. "Not X but Y" patterns
+6. Em dashes → hyphens
+7. Rule-of-three modifiers
+8. Cross-section template repetition
+9. Impersonal fragment stacking
+
+## STEP 6 — Cover Image
+- Read .stitch/design.md (6-slot layout)
+- Duplicate existing HTML from .stitch/designs/
+- Render: chrome-devtools → viewport 2400×1260×2 → screenshot
+- Downsample: magick → 2400×1260 → content/blog/<slug>/cover.png
+
+## STEP 7 — Validate
+- bin/hugo-build must pass (zero errors)
+- Chrome DevTools: zero console errors, zero 404s, og:image resolves
+
+## STEP 8 — Update Content Plan
+- Change topic status: 🔲 Planned → ✅ Published
+- Add commit SHA + changelog
+
+## STEP 9 — Commit
+- Stage: post directory + cover HTML + content plan update
+- Commit format: feat(content): add <topic> post
diff --git a/.apm/prompts/cover-image.prompt.md b/.apm/prompts/cover-image.prompt.md
new file mode 100644
index 000000000..2f53b5b4b
--- /dev/null
+++ b/.apm/prompts/cover-image.prompt.md
@@ -0,0 +1,19 @@
+# Cover Image Generation — Trigger Prompt
+#
+# Reusable prompt for generating blog post cover images via Stitch design system.
+# Full spec: .stitch/design.md (canonical 6-slot layout)
+# Full workflow: docs/workflows/cover-images.md
+
+Generate a cover image for a blog post.
+
+## Steps
+1. Read .stitch/design.md for the 6-slot layout specification
+2. Duplicate an existing HTML from .stitch/designs/ (e.g., rails-8-1-active-job-continuations-cover.html)
+3. Change ONLY: title, category, pill, 3 headline lines, 3 chip labels+values, status text
+4. Render: chrome-devtools new_page → emulate viewport=2400x1260x2 → wait 2s → take_screenshot
+5. Downsample: magick {screenshot} -filter Lanczos -resize 2400x1260 -sampling-factor 4:4:4 -strip -quality 95 content/blog/<slug>/cover.png
+
+## Design System
+- JetVelocity: obsidian dark, Ruby red (#cc342d), neon purple (#a855f7)
+- Font: Inter (headlines), JetBrains Mono (code/metrics)
+- 6-slot grid layout — see .stitch/design.md for exact positioning
diff --git a/.apm/prompts/research.prompt.md b/.apm/prompts/research.prompt.md
new file mode 100644
index 000000000..76d9c1507
--- /dev/null
+++ b/.apm/prompts/research.prompt.md
@@ -0,0 +1,26 @@
+# Research Protocol — Trigger Prompt
+#
+# Reusable prompt for research-first development.
+# MANDATORY sequence before any implementation.
+
+Research this topic before implementing: {TOPIC}
+
+## Research Sequence
+1. **Handbook search**: Search claude-context for existing patterns in /knowledge/ and docs/
+2. **Codebase search**: Search claude-context for similar implementations in this repo
+3. **Framework docs**: context7 for official documentation (Hugo, PostCSS, Ruby, etc.)
+4. **Package analysis**: package-search for dependency research (if using new packages)
+5. **Web search**: brave-search / searxng for current best practices (2025-2026)
+
+## Output Requirements
+- 5+ authoritative sources
+- Existing patterns found (or confirmation of none)
+- Framework documentation citations
+- Dependency evaluation (if applicable)
+- Recommendation with trade-offs
+
+## Anti-Patterns (BLOCKED)
+- Implementation before research
+- Single-source decisions
+- Dependencies installed without package-search verification
+- Slugs guessed without ls content/blog/ verification
diff --git a/.apm/skills/blog-post-pipeline/SKILL.md b/.apm/skills/blog-post-pipeline/SKILL.md
new file mode 100644
index 000000000..5f2324f46
--- /dev/null
+++ b/.apm/skills/blog-post-pipeline/SKILL.md
@@ -0,0 +1,41 @@
+# Blog Post Pipeline Skill
+#
+# Reusable 9-phase autonomous content pipeline for Hugo-based blogs.
+# This is a PRODUCER skill — published by jetthoughts/site for other repos.
+
+## Overview
+End-to-end pipeline from topic selection to published post with automated
+quality gates, multi-critic review, and cover image generation.
+
+## Phases
+1. Pick topic from content plan
+2. Read context (ICP, voice guide, brief)
+3. Research (5+ authoritative sources)
+4. Draft with voice enforcement + cadence quotas
+5. Review loop (3 parallel critic agents: founder, SEO/slop, editor)
+5c. Cold-eyes gate (9 explicit checks)
+6. Cover image (6-slot HTML → chrome-devtools render → lanczos downsample)
+7. Validate (Hugo build + Chrome DevTools)
+8. Update content plan
+9. Commit
+
+## Dependencies
+### MCP Servers
+- claude-context — semantic codebase search for existing posts
+- brave-search / searxng — web research
+- chrome-devtools — cover image rendering, validation
+
+### Other Skills
+- content-voice-enforcement — voice guide compliance
+- cover-image-generation — 6-slot layout rendering
+
+### Agents
+- content-creator — drafting
+- seo-specialist — SEO compliance
+- reviewer — quality gates
+
+## Trigger
+@.apm/prompts/blog-pipeline.prompt.md
+
+## Full Documentation
+docs/workflows/blog-pipeline.md
diff --git a/.apm/skills/content-voice-enforcement/SKILL.md b/.apm/skills/content-voice-enforcement/SKILL.md
new file mode 100644
index 000000000..df01df43f
--- /dev/null
+++ b/.apm/skills/content-voice-enforcement/SKILL.md
@@ -0,0 +1,37 @@
+# Content Voice Enforcement Skill
+#
+# Reusable voice enforcement for JetThoughts content — blog posts, LinkedIn,
+# and marketing copy targeting the ICP-E audience (non-technical founder
+# burned by a devshop).
+
+## Overview
+Enforces the JetThoughts voice guide, banned word list, structural patterns,
+cadence quotas, and quality tests on all content.
+
+## Voice Guide
+See .apm/instructions/voice-guide.instructions.md for the full compiled extract.
+Canonical source: docs/90-99-content-strategy/strategy-analysis/90.11-voice-guide.md
+
+## Two Voices
+- **Founder posts**: Rob Walling directness + Rand Fishkin vulnerability
+- **Tech posts**: Thoughtbot style (tension-first, opinionated, code as evidence)
+
+## Quality Tests (run on every paragraph)
+- **"Who" test**: who is doing what?
+- **"Show" test**: did I show the mechanic or just label it?
+- **"Coffee" test**: would a tired founder talk like this at a bar?
+- **"90/10" test**: ≥90% educational, ≤10% promotional
+- **"Trade-off" test**: acknowledged what the solution fails to do?
+- **"Practitioner" test**: replace generalizations with specific incidents
+
+## Slop Detector + Cold-Eyes Gate
+- Run slop-detector on all drafts
+- Run humanizer pass if flagged
+- Cold-eyes gate (9 checks) mandatory before publishing
+- See docs/workflows/blog-pipeline.md STEP 5a + STEP 5c
+
+## Full Documentation
+.apm/instructions/voice-guide.instructions.md
+.apm/instructions/icp-target.instructions.md
+docs/90-99-content-strategy/strategy-analysis/90.11-voice-guide.md
+docs/90-99-content-strategy/strategy-analysis/90.10-icp-primary-website-target.md
diff --git a/.apm/skills/cover-image-generation/SKILL.md b/.apm/skills/cover-image-generation/SKILL.md
new file mode 100644
index 000000000..a367207dc
--- /dev/null
+++ b/.apm/skills/cover-image-generation/SKILL.md
@@ -0,0 +1,32 @@
+# Cover Image Generation Skill
+#
+# Reusable skill for generating blog post cover images using the
+# JetVelocity 6-slot design system.
+
+## Overview
+Generates 2400×1260 cover images for blog posts using a 6-slot HTML layout
+rendered via chrome-devtools.
+
+## Design System
+- JetVelocity: obsidian dark, Ruby red (#cc342d), neon purple (#a855f7)
+- Font: Inter (headlines), JetBrains Mono (code/metrics)
+- 6-slot grid layout — canonical spec in .stitch/design.md
+
+## Workflow
+1. Read .stitch/design.md for the 6-slot layout specification
+2. Duplicate an existing HTML from .stitch/designs/
+3. Change ONLY: title, category, pill, 3 headline lines, 3 chip labels+values, status text
+4. Render: chrome-devtools new_page → emulate viewport=2400x1260x2 → wait 2s → take_screenshot
+5. Downsample: magick {screenshot} -filter Lanczos -resize 2400x1260 -sampling-factor 4:4:4 -strip -quality 95 content/blog/<slug>/cover.png
+6. Set frontmatter: cover_image: cover.png, cover_image_alt: descriptive alt text
+
+## Dependencies
+### MCP
+- chrome-devtools — rendering
+
+### System
+- ImageMagick (magick CLI) — downsampling
+
+## Full Documentation
+.stitch/design.md (canonical 6-slot spec)
+docs/workflows/cover-images.md
diff --git a/.apm/skills/css-bem-migration/SKILL.md b/.apm/skills/css-bem-migration/SKILL.md
new file mode 100644
index 000000000..9fcd1046f
--- /dev/null
+++ b/.apm/skills/css-bem-migration/SKILL.md
@@ -0,0 +1,49 @@
+# CSS BEM Migration Skill
+#
+# Reusable methodology for migrating from FL-Builder (Beaver Builder)
+# auto-generated classes to BEM component architecture.
+
+## Overview
+TDD-driven FL-node extraction: add BEM classes → verify visual fidelity →
+remove FL-nodes → retest. Micro-commits on every green test.
+
+## Pattern
+1. RED: Identify FL-node CSS dependencies
+2. GREEN: Add BEM classes to HTML, consolidate CSS
+3. REFACTOR: Remove duplicate FL-node CSS, cleanup
+4. TEST: `env PRECOMPILED_ASSETS=false bin/rake test:critical`
+5. COMMIT: Micro-commit on green tests
+
+## BEM Component Structure
+```html
+<!-- Before (FL-node) -->
+<div class="fl-col fl-node-yx43bujcaiqn hover-box">
+
+<!-- Transition (BEM + FL retention) -->
+<div class="fl-col fl-col-small hover-box c-feature-card">
+
+<!-- Final (Pure BEM) -->
+<div class="c-feature-card c-feature-card--variant">
+```
+
+## CSS Consolidation Pattern
+```css
+/* Before: 3 separate FL-node selectors */
+.fl-node-yx43bujcaiqn { width: 32%; margin-right: 30px; }
+.fl-node-ktz4ipj39vd6 { width: 36%; margin: 0 30px; }
+.fl-node-m39uvorzy5g8 { width: 32%; margin-left: 30px; }
+
+/* After: DRY BEM pattern */
+.c-feature-card { width: 32%; margin-right: 30px; }
+.c-feature-card--second { width: 36%; margin: 0 30px; }
+.c-feature-card--third { margin-right: 0; }
+```
+
+## Constraints
+- Never modify vendor files or dynamic templates
+- Preserve cascade order: Critical → Layout → Component → Theme → Footer
+- Run BOTH bin/test (macOS) AND bin/dtest (Linux Docker) before commit
+
+## Full Documentation
+docs/workflows/css-consolidation.md
+docs/projects/2509-css-migration/css-loading-order-analysis.md
diff --git a/.apm/skills/visual-regression-testing/SKILL.md b/.apm/skills/visual-regression-testing/SKILL.md
new file mode 100644
index 000000000..ff12576f9
--- /dev/null
+++ b/.apm/skills/visual-regression-testing/SKILL.md
@@ -0,0 +1,41 @@
+# Visual Regression Testing Skill
+#
+# Reusable visual regression testing protocol for Hugo static sites.
+
+## Overview
+Screenshot comparison-based visual testing using capybara-screenshot-diff.
+Blocks commits on any visual regression during refactoring.
+
+## Protocol
+1. Capture baseline screenshots BEFORE changes
+2. Make CSS/HTML changes
+3. Capture comparison screenshots AFTER changes
+4. Compare with tolerance:
+   - 0.0 for refactoring (zero visual changes)
+   - ≤0.03 for new features only
+5. Block commit on any difference > tolerance
+
+## Test Pattern
+```ruby
+def test_hero_section_visual_consistency
+  visit "/services/fractional-cto/"
+  screenshot_path = take_screenshot("fractional_cto_hero")
+  assert_visual_match(screenshot_path, "baseline/fractional_cto_hero.png",
+                      tolerance: 0.03)
+end
+```
+
+## Prohibited Patterns (ZERO TOLERANCE)
+- CSS class existence testing
+- HTML structure counting
+- Framework-generated selector usage
+- Method existence testing
+
+## Commands
+- `bin/test` — visual regression (macOS)
+- `bin/dtest` — visual regression (Linux/Docker, CI parity)
+- BOTH must pass before commit for HTML/CSS changes
+
+## Full Documentation
+.apm/instructions/tdd-enforcement.instructions.md
+docs/60.03-tdd-quality-enforcement.md
diff --git a/.claude/agents/analyst.md b/.claude/agents/analyst.md
new file mode 100644
index 000000000..dfa4ca692
--- /dev/null
+++ b/.claude/agents/analyst.md
@@ -0,0 +1,15 @@
+# analyst.agent.md — Thin Wrapper
+#
+# Full agent definitions: .claude/agents/core/analyst.md
+
+## Role
+Analysis agent. Performs research, root-cause analysis, and pattern discovery.
+
+## Instructions
+- @.apm/instructions/project-rules.instructions.md
+- @.apm/prompts/research.prompt.md
+
+## Research Protocol (MANDATORY)
+1. claude-context search FIRST
+2. Verify with rg/ls
+3. Read docs/workflows/BASE_HANDBOOK.md
diff --git a/.claude/agents/coder.md b/.claude/agents/coder.md
new file mode 100644
index 000000000..7c8b2615e
--- /dev/null
+++ b/.claude/agents/coder.md
@@ -0,0 +1,26 @@
+# coder.agent.md — Thin Wrapper
+#
+# Canonical instructions: .apm/instructions/project-rules.instructions.md
+# Full agent definitions: .claude/agents/core/coder.md
+# 
+# APM compiles this agent for all target tools. The canonical behavior
+# lives in the instruction files; this file pins the agent's role + pointers.
+
+## Role
+Primary implementation agent. Writes code following TDD cycle (RED → GREEN → REFACTOR).
+
+## Instructions
+- @.apm/instructions/project-rules.instructions.md
+- @.apm/instructions/tdd-enforcement.instructions.md
+
+## For CSS/HTML work
+- @.apm/instructions/css-consolidation.instructions.md
+
+## For Ruby work
+- @docs/workflows/ruby.md
+
+## Constraints
+- Behavior-focused tests ONLY — reject structural/CSS class tests
+- Run `bin/rake test:critical` after every micro-change
+- Never use `npm` — use `bun`
+- Never create ad hoc `*.sh` test scripts
diff --git a/.claude/agents/content-marketer.md b/.claude/agents/content-marketer.md
new file mode 100644
index 000000000..053ecd0a7
--- /dev/null
+++ b/.claude/agents/content-marketer.md
@@ -0,0 +1,10 @@
+# content-marketer.agent.md — Thin Wrapper
+#
+# Full agent definitions: .claude/agents/content/content-marketer.md
+
+## Role
+Content marketing agent. Plans content calendars, researches topics, audits SEO performance.
+
+## Instructions
+- @.apm/instructions/icp-target.instructions.md
+- @docs/projects/2510-seo-content-strategy/20-29-strategy/20.07-content-plan-icp-e-q2-2026.md
diff --git a/.claude/agents/planner.md b/.claude/agents/planner.md
new file mode 100644
index 000000000..837cedb13
--- /dev/null
+++ b/.claude/agents/planner.md
@@ -0,0 +1,9 @@
+# planner.agent.md — Thin Wrapper
+#
+# Full agent definitions: .claude/agents/core/planner.md
+
+## Role
+Planning agent. Breaks down complex tasks into ordered steps, writes todo lists, and sequences work.
+
+## Instructions
+- @.apm/instructions/project-rules.instructions.md
diff --git a/.claude/agents/qa-browser-tester.md b/.claude/agents/qa-browser-tester.md
new file mode 100644
index 000000000..3a18ca639
--- /dev/null
+++ b/.claude/agents/qa-browser-tester.md
@@ -0,0 +1,16 @@
+# qa-browser-tester.agent.md — Thin Wrapper
+#
+# Full agent definitions: .claude/agents/validation/qa-browser-tester.md
+
+## Role
+QA browser tester. Runs Chrome DevTools validation after HTML/CSS/JS changes.
+
+## Instructions
+- @.apm/instructions/project-rules.instructions.md
+- @.apm/instructions/tdd-enforcement.instructions.md
+
+## Gates
+- Console: ZERO JS errors
+- Network: ZERO 404s for assets
+- Capture desktop + mobile screenshots
+- All gates pass → commit
diff --git a/.claude/agents/researcher.md b/.claude/agents/researcher.md
new file mode 100644
index 000000000..5d953465d
--- /dev/null
+++ b/.claude/agents/researcher.md
@@ -0,0 +1,10 @@
+# researcher.agent.md — Thin Wrapper
+#
+# Full agent definitions: .claude/agents/core/researcher.md
+
+## Role
+Research agent. Discovers existing patterns, framework docs, and best practices before implementation.
+
+## Instructions
+- @.apm/instructions/project-rules.instructions.md
+- @.apm/prompts/research.prompt.md
diff --git a/.claude/agents/reviewer.md b/.claude/agents/reviewer.md
new file mode 100644
index 000000000..989624346
--- /dev/null
+++ b/.claude/agents/reviewer.md
@@ -0,0 +1,17 @@
+# reviewer.agent.md — Thin Wrapper
+#
+# Canonical instructions: .apm/instructions/project-rules.instructions.md
+# Full agent definitions: .claude/agents/core/reviewer.md
+
+## Role
+Code review agent. Validates correctness, adherence to project rules, and TDD compliance.
+
+## Instructions
+- @.apm/instructions/project-rules.instructions.md
+- @.apm/instructions/tdd-enforcement.instructions.md
+
+## Gates
+- Zero structural test patterns (CSS class testing, HTML counting)
+- Visual regression tolerance verified
+- Four-eyes principle enforced for all critical changes
+- Feature branch + PR pattern for HTML/CSS/template changes
diff --git a/.claude/agents/ruby-coder.md b/.claude/agents/ruby-coder.md
new file mode 100644
index 000000000..510d186df
--- /dev/null
+++ b/.claude/agents/ruby-coder.md
@@ -0,0 +1,15 @@
+# ruby-coder.agent.md — Thin Wrapper
+#
+# Full agent definitions: .claude/agents/ruby/ruby-coder.md
+
+## Role
+Ruby implementation agent. Writes Ruby code (tests, sync scripts, lib/) following TDD.
+
+## Instructions
+- @.apm/instructions/project-rules.instructions.md
+- @.apm/instructions/tdd-enforcement.instructions.md
+- @docs/workflows/ruby.md
+
+## Commands
+- `bin/rake test:critical`
+- `bundle exec standardrb --fix`
diff --git a/.claude/agents/ruby-reviewer.md b/.claude/agents/ruby-reviewer.md
new file mode 100644
index 000000000..bdfcf27b8
--- /dev/null
+++ b/.claude/agents/ruby-reviewer.md
@@ -0,0 +1,14 @@
+# ruby-reviewer.agent.md — Thin Wrapper
+#
+# Full agent definitions: .claude/agents/ruby/ruby-reviewer.md
+
+## Role
+Ruby code reviewer. Validates Ruby code quality, TDD compliance, and Minitest patterns.
+
+## Instructions
+- @.apm/instructions/tdd-enforcement.instructions.md
+
+## Constraints
+- Behavior-focused tests ONLY — reject structural testing
+- StandardRB compliance
+- No ad hoc test scripts
diff --git a/.claude/agents/screenshot-guardian.md b/.claude/agents/screenshot-guardian.md
new file mode 100644
index 000000000..f160b3b75
--- /dev/null
+++ b/.claude/agents/screenshot-guardian.md
@@ -0,0 +1,15 @@
+# screenshot-guardian.agent.md — Thin Wrapper
+#
+# Full agent definitions: .claude/agents/core/screenshot-guardian.md
+
+## Role
+Visual regression guardian. Blocking agent that prevents test masking (updating baselines during failures).
+
+## Instructions
+- @.apm/instructions/tdd-enforcement.instructions.md
+- @.claude/hooks/pre-commit-screenshot-validation.md
+
+## Gates
+- ZERO tolerance: screenshot updates during visual test failures
+- Tolerance: 0.0 for refactors, ≤0.03 for new features
+- Both bin/test AND bin/dtest must pass before commit
diff --git a/.claude/agents/tester.md b/.claude/agents/tester.md
new file mode 100644
index 000000000..8bc5f213b
--- /dev/null
+++ b/.claude/agents/tester.md
@@ -0,0 +1,25 @@
+# tester.agent.md — Thin Wrapper
+#
+# Canonical instructions: .apm/instructions/project-rules.instructions.md
+# Full agent definitions: .claude/agents/core/tester.md
+
+## Role
+Test agent. Writes and runs tests following TDD methodology.
+
+## Instructions
+- @.apm/instructions/project-rules.instructions.md
+- @.apm/instructions/tdd-enforcement.instructions.md
+
+## Framework
+- Minitest (test/system/, test/unit/)
+- capybara-screenshot-diff for visual regression
+
+## Commands
+- `bin/rake test:critical` — critical tests
+- `bin/test` — visual regression (macOS)
+- `bin/dtest` — visual regression (Linux/Docker)
+
+## Constraints
+- Behavior-focused tests ONLY
+- Assert shape, not config values
+- Run both bin/test AND bin/dtest before commit for HTML/CSS changes
diff --git a/.claude/agents/tutorial-engineer.md b/.claude/agents/tutorial-engineer.md
new file mode 100644
index 000000000..14374cb5e
--- /dev/null
+++ b/.claude/agents/tutorial-engineer.md
@@ -0,0 +1,10 @@
+# tutorial-engineer.agent.md — Thin Wrapper
+#
+# Full agent definitions: .claude/agents/content/tutorial-engineer.md
+
+## Role
+Tutorial engineer. Creates step-by-step how-to content, validates code examples, and ensures copy-paste-ability.
+
+## Instructions
+- @.apm/instructions/voice-guide.instructions.md
+- @.apm/instructions/icp-target.instructions.md
diff --git a/.claude/agents/ui-problem-diagnosis.md b/.claude/agents/ui-problem-diagnosis.md
new file mode 100644
index 000000000..9f88e69ee
--- /dev/null
+++ b/.claude/agents/ui-problem-diagnosis.md
@@ -0,0 +1,10 @@
+# ui-problem-diagnosis.agent.md — Thin Wrapper
+#
+# Full agent definitions: .claude/agents/validation/ui-problem-diagnosis-specialist.md
+
+## Role
+UI problem diagnosis specialist. Investigates visual regressions, cross-browser issues, and CSS bugs.
+
+## Instructions
+- @.apm/instructions/css-consolidation.instructions.md
+- @docs/projects/2509-css-migration/css-loading-order-analysis.md
diff --git a/.claude/agents/ux-browser-validator.md b/.claude/agents/ux-browser-validator.md
new file mode 100644
index 000000000..118ea4ffa
--- /dev/null
+++ b/.claude/agents/ux-browser-validator.md
@@ -0,0 +1,14 @@
+# ux-browser-validator.agent.md — Thin Wrapper
+#
+# Full agent definitions: .claude/agents/validation/ux-browser-validator.md
+
+## Role
+UX browser validator. Validates UX interactions, accessibility, and cross-browser behavior.
+
+## Instructions
+- @.apm/instructions/project-rules.instructions.md
+
+## Gates
+- WCAG 2.1 AA accessibility compliance
+- Semantic HTML elements (h1, button, main, nav)
+- Responsive design: desktop + mobile verification
diff --git a/.claude/commands/blog-pipeline.md b/.claude/commands/blog-pipeline.md
new file mode 100644
index 000000000..5c9dae536
--- /dev/null
+++ b/.claude/commands/blog-pipeline.md
@@ -0,0 +1,79 @@
+---
+{}
+---
+
+# Blog Post Pipeline — Trigger Prompt
+#
+# Copy this prompt to write the next blog post from the content plan.
+# Full pipeline: docs/workflows/blog-pipeline.md
+#
+# This is a reusable PROMPT — APM compiles it to slash-commands for
+# each AI tool (e.g., /blog-post in Claude, #blog-post in Copilot).
+
+Write the next blog post following the full pipeline.
+
+## STEP 1 — Pick Topic
+Read the content plan: docs/projects/2510-seo-content-strategy/20-29-strategy/20.07-content-plan-icp-e-q2-2026.md
+Find the next 🔲 Planned topic. Confirm slug, primary keyword, target publish date.
+
+## STEP 2 — Read Context (mandatory, do not skip)
+- ICP: docs/90-99-content-strategy/strategy-analysis/90.10-icp-primary-website-target.md
+- Voice guide: docs/90-99-content-strategy/strategy-analysis/90.11-voice-guide.md
+- Content plan brief: read the topic's brief section
+- Design spec: .stitch/design.md
+
+## STEP 3 — Research
+- Web search for 5+ authoritative sources
+- Check existing posts for duplication: ls content/blog/ | grep -iE "<key-terms>"
+- Verify every statistic has a citable source
+
+## STEP 3b — Find Internal Posts to Reference
+- Use claude-context MCP search FIRST
+- For tag/slug lookups: docs/blog-post-index.md
+- Verify each slug: ls content/blog/<slug>/index.md
+- Collect at least 4 verified slugs before drafting
+
+## STEP 4 — Draft
+Create content/blog/<slug>/index.md with:
+- Frontmatter: title ≤45 chars, description 150-160 chars, date, slug, keywords, tags
+- Voice: plain English, empathy first, founder-to-founder tone
+- Structure: punchy hook → keyword in first 100 words → numbered H2s → "When NOT to" → JT proof → CTA → related reading → further reading
+- 4+ internal links in body, 5+ external citations
+- Paragraphs ≤3 sentences
+- Cut test: "what does the reader DO differently?" — delete if answer is "nothing new"
+
+## STEP 5 — Review Loop
+Spawn 3 parallel critic agents:
+1. Founder persona (Alex, burned CEO) — ≥8/10
+2. SEO/slop detector — ≥8/10 humanization
+3. Senior copy editor — "publish as-is" or "minor edits"
+
+## STEP 5c — Cold-Eyes Gate (mandatory)
+Spawn fresh-context reviewer with 9 explicit checks:
+1. Frontmatter vs body consistency
+2. Fluffy AI narration detection
+3. Fake authority stamps
+4. Timeline fabrication
+5. "Not X but Y" patterns
+6. Em dashes → hyphens
+7. Rule-of-three modifiers
+8. Cross-section template repetition
+9. Impersonal fragment stacking
+
+## STEP 6 — Cover Image
+- Read .stitch/design.md (6-slot layout)
+- Duplicate existing HTML from .stitch/designs/
+- Render: chrome-devtools → viewport 2400×1260×2 → screenshot
+- Downsample: magick → 2400×1260 → content/blog/<slug>/cover.png
+
+## STEP 7 — Validate
+- bin/hugo-build must pass (zero errors)
+- Chrome DevTools: zero console errors, zero 404s, og:image resolves
+
+## STEP 8 — Update Content Plan
+- Change topic status: 🔲 Planned → ✅ Published
+- Add commit SHA + changelog
+
+## STEP 9 — Commit
+- Stage: post directory + cover HTML + content plan update
+- Commit format: feat(content): add <topic> post
\ No newline at end of file
diff --git a/.claude/commands/cover-image.md b/.claude/commands/cover-image.md
new file mode 100644
index 000000000..024cd8573
--- /dev/null
+++ b/.claude/commands/cover-image.md
@@ -0,0 +1,23 @@
+---
+{}
+---
+
+# Cover Image Generation — Trigger Prompt
+#
+# Reusable prompt for generating blog post cover images via Stitch design system.
+# Full spec: .stitch/design.md (canonical 6-slot layout)
+# Full workflow: docs/workflows/cover-images.md
+
+Generate a cover image for a blog post.
+
+## Steps
+1. Read .stitch/design.md for the 6-slot layout specification
+2. Duplicate an existing HTML from .stitch/designs/ (e.g., rails-8-1-active-job-continuations-cover.html)
+3. Change ONLY: title, category, pill, 3 headline lines, 3 chip labels+values, status text
+4. Render: chrome-devtools new_page → emulate viewport=2400x1260x2 → wait 2s → take_screenshot
+5. Downsample: magick {screenshot} -filter Lanczos -resize 2400x1260 -sampling-factor 4:4:4 -strip -quality 95 content/blog/<slug>/cover.png
+
+## Design System
+- JetVelocity: obsidian dark, Ruby red (#cc342d), neon purple (#a855f7)
+- Font: Inter (headlines), JetBrains Mono (code/metrics)
+- 6-slot grid layout — see .stitch/design.md for exact positioning
\ No newline at end of file
diff --git a/.claude/commands/research.md b/.claude/commands/research.md
new file mode 100644
index 000000000..59101394b
--- /dev/null
+++ b/.claude/commands/research.md
@@ -0,0 +1,30 @@
+---
+{}
+---
+
+# Research Protocol — Trigger Prompt
+#
+# Reusable prompt for research-first development.
+# MANDATORY sequence before any implementation.
+
+Research this topic before implementing: {TOPIC}
+
+## Research Sequence
+1. **Handbook search**: Search claude-context for existing patterns in /knowledge/ and docs/
+2. **Codebase search**: Search claude-context for similar implementations in this repo
+3. **Framework docs**: context7 for official documentation (Hugo, PostCSS, Ruby, etc.)
+4. **Package analysis**: package-search for dependency research (if using new packages)
+5. **Web search**: brave-search / searxng for current best practices (2025-2026)
+
+## Output Requirements
+- 5+ authoritative sources
+- Existing patterns found (or confirmation of none)
+- Framework documentation citations
+- Dependency evaluation (if applicable)
+- Recommendation with trade-offs
+
+## Anti-Patterns (BLOCKED)
+- Implementation before research
+- Single-source decisions
+- Dependencies installed without package-search verification
+- Slugs guessed without ls content/blog/ verification
\ No newline at end of file
diff --git a/.claude/rules/css-consolidation.md b/.claude/rules/css-consolidation.md
new file mode 100644
index 000000000..0f1faae2d
--- /dev/null
+++ b/.claude/rules/css-consolidation.md
@@ -0,0 +1,30 @@
+# CSS Consolidation Rules — JetThoughts Site
+#
+# MANDATORY for all CSS/HTML template changes.
+# Full docs: docs/workflows/css-consolidation.md
+# Project: docs/projects/2509-css-migration/css-loading-order-analysis.md
+
+## Architecture
+- **Methodology**: BEM (Block__Element--Modifier)
+- **Pipeline**: PostCSS (import, mixins, nested, autoprefixer, cssnano, purgecss)
+- **Cascade order**: Critical → Layout → Component → Theme → Footer
+
+## Rules
+- Never modify vendor files (`css/vendors/*.min.css`) or dynamic templates (`css/dynamic-*.css`)
+- Preserve cascade order in all changes
+- Extract whole rule sets; keep page-specific `.fl-node-{hash}` in original files
+- Run `bin/rake test:critical` after each extraction
+
+## FL-Node Extraction (Beaver Builder → BEM)
+- **Phase**: TDD RED-GREEN-REFACTOR cycle
+- **Validation**: `env PRECOMPILED_ASSETS=false bin/rake test:critical` after each change
+- **Pattern**: Add BEM classes first → verify visual fidelity → remove FL-nodes → retest
+- **Commit**: micro-commit on green tests only
+
+## Visual Regression
+- Tolerance: 0.0 for refactoring (zero visual changes)
+- Tolerance: ≤0.03 for new features only
+- Run BOTH `bin/test` and `bin/dtest` before commit
+- Visual failures are commit blockers, not warnings
+
+@see docs/workflows/css-consolidation.md
diff --git a/.claude/rules/icp-target.md b/.claude/rules/icp-target.md
new file mode 100644
index 000000000..e9457d522
--- /dev/null
+++ b/.claude/rules/icp-target.md
@@ -0,0 +1,46 @@
+# ICP Target — JetThoughts Primary Website Audience
+#
+# CANONICAL SOURCE: docs/90-99-content-strategy/strategy-analysis/90.10-icp-primary-website-target.md
+# This file is a COMPILED EXTRACT for APM — the full doc is authoritative.
+# Update the canonical source first, then sync this extract.
+#
+# MANDATORY reading before content, design, marketing, or blog work.
+# Compiled by APM into every AI tool's context.
+
+## ICP-E: Non-Technical Founder Burned by a Devshop
+
+### Who They Are
+- 35-50 years old
+- Non-technical or semi-technical
+- Raised funding ($500K–$2M seed)
+- Hired a dev shop or freelancers
+- Missed deadlines, buggy product, mounting technical debt
+- Starting to question: "Is it me? Is it them? How do I know?"
+
+### Their Pain
+- Can't evaluate code quality themselves
+- Dev shop says "almost done" for months
+- Every new feature breaks something else
+- No tests, no documentation, no deploy pipeline
+- Starting to lose investor confidence
+
+### What They Need
+- Someone who speaks plain English, not jargon
+- Proof that the problem is real and solvable (specific stories, not vague promises)
+- A clear path out: audit → stabilize → build right
+- Trust signals: named clients, specific results, honest trade-offs
+
+### What Resonates
+- Stories of other founders in the same situation
+- Specific numbers: "$40K wasted," "11-minute outage," "84 support tickets"
+- Honest acknowledgment that rescue is hard and expensive
+- The idea that good engineering is a competitive advantage
+- "You're not crazy — this IS what bad engineering looks like"
+
+### What Alienates
+- jargon-dense technical explanations before establishing trust
+- "We're the best" claims without evidence
+- Generic consulting language ("partner with you," "drive outcomes")
+- Ignoring the emotional reality (shame, frustration, fear of looking stupid)
+
+@see docs/90-99-content-strategy/strategy-analysis/90.10-icp-primary-website-target.md
diff --git a/.claude/rules/project-rules.md b/.claude/rules/project-rules.md
new file mode 100644
index 000000000..deb7313e4
--- /dev/null
+++ b/.claude/rules/project-rules.md
@@ -0,0 +1,79 @@
+# JetThoughts Site — Project Rules
+#
+# This is the SINGLE SOURCE for AI agent instructions across all tools.
+# APM compiles this to each tool's expected path (CLAUDE.md, .github/
+# copilot-instructions.md, etc.). Edit here, not in tool-specific files.
+#
+# CANONICAL SOURCE: THIS FILE. The existing CLAUDE.md is a Claude-specific
+# superset; over time CLI tool instructions converge here. Voice, ICP, TDD,
+# and CSS rules are split into separate instruction files for modularity.
+#
+# Full workflow docs: docs/workflows/
+# Full content docs: docs/90-99-content-strategy/
+
+## Stack
+- **Site**: Hugo 0.147.9 Extended static site generator
+- **Theme**: Custom "Beaver" theme (BEM + PostCSS)
+- **Tests**: Ruby 3.4, Minitest, Capybara, Selenium WebDriver
+- **CSS**: PostCSS pipeline (autoprefixer, cssnano, purgecss)
+- **Package managers**: Bun (JS, NOT npm), Bundler (Ruby)
+- **CI/CD**: GitHub Actions → GitHub Pages
+
+## Build & Test Commands
+- Build: `bin/hugo-build`
+- Test (critical): `bin/rake test:critical`
+- Test (visual/macOS): `bin/test`
+- Test (visual/Linux Docker): `bin/dtest`
+- Lint Ruby: `bundle exec standardrb --fix`
+
+## Research Protocol (MANDATORY — before any change)
+1. Search codebase with `claude-context` MCP FIRST
+2. Verify exact filenames with `rg`/`ls` as fallback
+3. Read `docs/workflows/BASE_HANDBOOK.md` at session start
+4. Read `docs/workflows/flow-router.md` to route to correct workflow
+5. For blog content: find existing posts via `docs/blog-post-index.md`
+6. Never guess slugs — verify with `ls content/blog/<slug>/index.md`
+
+## Non-Negotiables
+- **Zero generic AI language** in content — all AI-sounding phrases rejected
+- **Zero Hugo build breaks** — `bin/hugo-build` must pass for content changes
+- **Zero unsupported claims** — all assertions must have citations
+- **No ad hoc test scripts** — use `bin/rake test:critical` and `bin/test`
+- **No Python for analysis** — prefer `rg`, `sed`, `awk`, shell tools
+- **No duplicate throwaway files** — no `*_refactored.*`, `*_new.*`, `*_backup.*`
+- **New docs only under `docs/workflows/`** — else edit existing files
+- **Never commit coordinator/verification reports** — write to `/tmp/` instead
+
+## Content Work (Blog Posts, LinkedIn, Pages)
+- **BEFORE writing**: Read voice guide + ICP target (see instructions/voice-guide, instructions/icp-target)
+- **Blog posts**: Follow `docs/workflows/blog-pipeline.md` end-to-end
+- **LinkedIn**: Follow `docs/workflows/linkedin-post-pipeline.md`
+- **Cover images**: Follow `docs/workflows/cover-images.md` + `.stitch/design.md`
+- Run slop-detector + humanizer pass on all drafts
+- Cold-eyes gate (9 checks) mandatory before publishing
+
+## CSS / HTML Changes
+- **BEM methodology** mandatory for all new CSS
+- **FL-node extraction**: TDD cycle — RED → GREEN → REFACTOR → TEST
+- **Visual regression tolerance**: 0.0 for refactors, ≤0.03 for new features
+- **MANDATORY**: Run BOTH `bin/test` (macOS) AND `bin/dtest` (Linux Docker) before commit
+- **Feature branches**: HTML/CSS/template changes → feature branch → one PR per sprint
+- Zero tolerance: CSS class existence testing, framework selector usage in tests
+
+## TDD & Testing
+- **Cycle**: RED → GREEN → REFACTOR
+- **Framework**: Minitest (test/system/, test/unit/)
+- **Behavior-focused tests ONLY** — reject implementation/existence/config tests
+- **Avoid fragile config assertions** — assert shape, not tunable values
+- **Visual regression**: screenshot comparison, 0.0 tolerance for refactoring
+- Run `bin/rake test:critical` after every micro-change (<10 lines)
+- Chrome DevTools validation after HTML/CSS/JS changes: zero console errors, zero 404s
+
+## Behavioral Constraints
+- **User says "code is bad" or "over-engineered"**: HALT, 5-Why analysis, fix config, proceed
+- **Reflection triggers**: zero-defect tracking, test masking detection
+- **Four-eyes principle**: paired agents for all critical changes
+- **Direct-to-master only for**: content-only blog edits, commit fixes, docs, CLAUDE.md policy updates
+
+@see docs/workflows/ for full workflow documentation
+@see docs/90-99-content-strategy/ for content strategy
diff --git a/.claude/rules/tdd-enforcement.md b/.claude/rules/tdd-enforcement.md
new file mode 100644
index 000000000..a532cfa19
--- /dev/null
+++ b/.claude/rules/tdd-enforcement.md
@@ -0,0 +1,62 @@
+# TDD Enforcement — JetThoughts Site
+#
+# MANDATORY for all code changes. Extracted from CLAUDE.md TDD section.
+# Full docs: docs/60.03-tdd-quality-enforcement.md, docs/60.04-four-eyes-principle.md
+
+## TDD Cycle
+1. **RED**: Write a failing test that validates BEHAVIOR (not implementation)
+2. **GREEN**: Minimal code to make the test pass
+3. **REFACTOR**: Clean up while tests stay green
+4. **COMMIT**: Micro-commit on green tests
+
+## Test Quality Rules (ZERO TOLERANCE)
+
+### Prohibited Test Patterns — Agent MUST Reject
+- ❌ CSS class existence testing (`.fl-node-*`, `.has_css?(".px-4")`)
+- ❌ HTML structure counting (`page.all("div").count`)
+- ❌ Framework-generated selector usage (FL-Builder, Tailwind utility classes)
+- ❌ Method existence testing (`assert SomeClass.method_defined?`)
+- ❌ Config value assertions (hardcoded `q=90`, exact file sizes, specific dimensions)
+- ❌ Implementation detail testing
+
+### Required Test Patterns
+- ✅ Visual regression: `take_screenshot` → `assert_visual_match` with tolerance
+- ✅ User-visible content: `assert page.has_content?("text")`
+- ✅ Semantic elements: `assert page.has_css?("h1", text: "...")`
+- ✅ User interaction: `click_button` → `assert_current_path`
+- ✅ Shape assertions: `q=\d+`, `src contains wsrv.nl`, not `q=90`, not exact URLs
+
+### Avoid Fragile Config Assertions
+Don't hardcode tunable values. Assert the *shape*, not the configuration:
+- Bad: `assert_match /q=90/, src`
+- Good: `assert_match /q=\d+/, src`
+- Bad: `assert_match /w=360/, src`
+- Good: `assert has_css?("picture source")`
+
+## Visual Regression Gates (BLOCKING for HTML/CSS changes)
+- **Tolerance**: 0.0 for refactoring, ≤0.03 for new features
+- **Protocol**: Capture baseline BEFORE changes → compare AFTER → block on any diff >0%
+- **MANDATORY**: Run BOTH `bin/test` (macOS) AND `bin/dtest` (Linux Docker) before commit
+- Visual regression failures are COMMIT BLOCKERS, not warnings
+
+## Test Runner
+- Framework: Minitest
+- Critical tests: `bin/rake test:critical` (after every micro-change <10 lines)
+- Visual macOS: `bin/test`
+- Visual Linux/Docker: `bin/dtest`
+- NEVER create ad hoc `*.sh` test scripts
+
+## Chrome DevTools Validation (MANDATORY after HTML/CSS/JS)
+1. Start Hugo dev server, open page in Chrome DevTools
+2. Console: ZERO JS errors
+3. Network: ZERO 404s
+4. Capture desktop + mobile screenshots
+5. All gates pass → commit. Any gate fails → fix, repeat
+
+## Feature Branch + PR (HTML/CSS/template changes)
+1. `git checkout -b <sprint-name>`
+2. Multiple related commits on branch
+3. Both test gates green on each commit
+4. `git push -u origin <branch>`
+5. ONE PR per sprint via `gh pr create`
+6. Direct-to-master only for: content-only blog edits, commit fixes, docs, CLAUDE.md updates
diff --git a/.claude/rules/voice-guide.md b/.claude/rules/voice-guide.md
new file mode 100644
index 000000000..97aeec148
--- /dev/null
+++ b/.claude/rules/voice-guide.md
@@ -0,0 +1,68 @@
+# Voice Guide — JetThoughts Content
+#
+# CANONICAL SOURCE: docs/90-99-content-strategy/strategy-analysis/90.11-voice-guide.md
+# This file is a COMPILED EXTRACT for APM — the full doc is authoritative.
+# Update the canonical source first, then sync this extract.
+#
+# MANDATORY reading before writing ANY blog post, LinkedIn post, or marketing copy.
+# Compiled by APM into every AI tool's context.
+
+## Two Voices
+
+### Founder Posts (Paul Keen / JT leadership)
+- **Rob Walling directness** + **Rand Fishkin vulnerability**
+- Plain English, empathy, specific numbers
+- Talk to a non-technical founder burned by a devshop
+- "Been there" energy — practitioner not vendor
+- Trade-offs acknowledged honestly
+
+### Tech Posts (engineering team)
+- **Thoughtbot style**: open with tension, not features
+- Own your opinions — no "it depends" hedging
+- Code as narrative evidence, not decoration
+- Honest trade-offs: what does this solution fail to do?
+
+## Banned Words
+unlock, harness, leverage, dive in, game-changer, journey, seamless, robust,
+supercharge, revolutionize, embark, delve, discover, blast radius, suspect list
+
+## Banned Structural Patterns
+- Rule of three (parallel triads)
+- Signposting ("This post is about...", "Here's what you need to know")
+- Bold inline-header lists
+- Triple rhetorical questions
+- Slogany parallel closings ("X is A. Y is B.")
+- Therapist voice ("That's the hardest part")
+- Copula avoidance ("serves as" → use "is")
+- Impersonal fragment stacking — every sentence needs a human subject
+- Sustained staccato (3+ consecutive short fragments or uniform-length sentences)
+- Command structure repetition (3+ paragraphs starting with imperatives)
+- Apologetic caveats ("every project is different")
+- Telling instead of showing — replace adjectives with concrete mechanics
+- Fake authority generalizations ("We've seen this on every codebase")
+- Timeline fabrication — verify technology release dates before using "last year"
+- Use `-` not `—` for all dashes
+
+## Quality Tests (run on every paragraph)
+- **"Who" test**: who is doing what? If subject is a thing, rewrite with a person
+- **"Show" test**: did I show the mechanic or just label it?
+- **"Coffee" test**: would a tired founder talk like this at a bar?
+- **"90/10" test**: is ≥90% educational? Is promotion ≤10%?
+- **"Trade-off" test**: did I acknowledge what the solution fails to do?
+- **"Practitioner" test**: replace generalized scenarios with specific incidents
+
+## Mannerism Noun Cap
+discipline, cycle, pattern, canonical, workflow, framework, cadence, rhythm,
+contract, gate, gatekeeper, layer, round, review, pass
+→ Max 5 occurrences total per post. Vary to the concrete thing happening.
+
+## Cadence Quotas (per H2 section)
+- At least one sentence <9 words AND one >22 words
+- No 3 consecutive sentences within 3 words of each other
+- One single-sentence paragraph per H2
+- Max 3 sentences / 70 words per paragraph
+- No two sentences in same paragraph starting with same first word
+- "The..." paragraph openers: <20% of total paragraphs
+- "We..." paragraph openers: <15% of total paragraphs
+
+@see docs/90-99-content-strategy/strategy-analysis/90.11-voice-guide.md
diff --git a/.claude/skills/blog-post-pipeline/SKILL.md b/.claude/skills/blog-post-pipeline/SKILL.md
new file mode 100644
index 000000000..5f2324f46
--- /dev/null
+++ b/.claude/skills/blog-post-pipeline/SKILL.md
@@ -0,0 +1,41 @@
+# Blog Post Pipeline Skill
+#
+# Reusable 9-phase autonomous content pipeline for Hugo-based blogs.
+# This is a PRODUCER skill — published by jetthoughts/site for other repos.
+
+## Overview
+End-to-end pipeline from topic selection to published post with automated
+quality gates, multi-critic review, and cover image generation.
+
+## Phases
+1. Pick topic from content plan
+2. Read context (ICP, voice guide, brief)
+3. Research (5+ authoritative sources)
+4. Draft with voice enforcement + cadence quotas
+5. Review loop (3 parallel critic agents: founder, SEO/slop, editor)
+5c. Cold-eyes gate (9 explicit checks)
+6. Cover image (6-slot HTML → chrome-devtools render → lanczos downsample)
+7. Validate (Hugo build + Chrome DevTools)
+8. Update content plan
+9. Commit
+
+## Dependencies
+### MCP Servers
+- claude-context — semantic codebase search for existing posts
+- brave-search / searxng — web research
+- chrome-devtools — cover image rendering, validation
+
+### Other Skills
+- content-voice-enforcement — voice guide compliance
+- cover-image-generation — 6-slot layout rendering
+
+### Agents
+- content-creator — drafting
+- seo-specialist — SEO compliance
+- reviewer — quality gates
+
+## Trigger
+@.apm/prompts/blog-pipeline.prompt.md
+
+## Full Documentation
+docs/workflows/blog-pipeline.md
diff --git a/.claude/skills/content-voice-enforcement/SKILL.md b/.claude/skills/content-voice-enforcement/SKILL.md
new file mode 100644
index 000000000..df01df43f
--- /dev/null
+++ b/.claude/skills/content-voice-enforcement/SKILL.md
@@ -0,0 +1,37 @@
+# Content Voice Enforcement Skill
+#
+# Reusable voice enforcement for JetThoughts content — blog posts, LinkedIn,
+# and marketing copy targeting the ICP-E audience (non-technical founder
+# burned by a devshop).
+
+## Overview
+Enforces the JetThoughts voice guide, banned word list, structural patterns,
+cadence quotas, and quality tests on all content.
+
+## Voice Guide
+See .apm/instructions/voice-guide.instructions.md for the full compiled extract.
+Canonical source: docs/90-99-content-strategy/strategy-analysis/90.11-voice-guide.md
+
+## Two Voices
+- **Founder posts**: Rob Walling directness + Rand Fishkin vulnerability
+- **Tech posts**: Thoughtbot style (tension-first, opinionated, code as evidence)
+
+## Quality Tests (run on every paragraph)
+- **"Who" test**: who is doing what?
+- **"Show" test**: did I show the mechanic or just label it?
+- **"Coffee" test**: would a tired founder talk like this at a bar?
+- **"90/10" test**: ≥90% educational, ≤10% promotional
+- **"Trade-off" test**: acknowledged what the solution fails to do?
+- **"Practitioner" test**: replace generalizations with specific incidents
+
+## Slop Detector + Cold-Eyes Gate
+- Run slop-detector on all drafts
+- Run humanizer pass if flagged
+- Cold-eyes gate (9 checks) mandatory before publishing
+- See docs/workflows/blog-pipeline.md STEP 5a + STEP 5c
+
+## Full Documentation
+.apm/instructions/voice-guide.instructions.md
+.apm/instructions/icp-target.instructions.md
+docs/90-99-content-strategy/strategy-analysis/90.11-voice-guide.md
+docs/90-99-content-strategy/strategy-analysis/90.10-icp-primary-website-target.md
diff --git a/.claude/skills/cover-image-generation/SKILL.md b/.claude/skills/cover-image-generation/SKILL.md
new file mode 100644
index 000000000..a367207dc
--- /dev/null
+++ b/.claude/skills/cover-image-generation/SKILL.md
@@ -0,0 +1,32 @@
+# Cover Image Generation Skill
+#
+# Reusable skill for generating blog post cover images using the
+# JetVelocity 6-slot design system.
+
+## Overview
+Generates 2400×1260 cover images for blog posts using a 6-slot HTML layout
+rendered via chrome-devtools.
+
+## Design System
+- JetVelocity: obsidian dark, Ruby red (#cc342d), neon purple (#a855f7)
+- Font: Inter (headlines), JetBrains Mono (code/metrics)
+- 6-slot grid layout — canonical spec in .stitch/design.md
+
+## Workflow
+1. Read .stitch/design.md for the 6-slot layout specification
+2. Duplicate an existing HTML from .stitch/designs/
+3. Change ONLY: title, category, pill, 3 headline lines, 3 chip labels+values, status text
+4. Render: chrome-devtools new_page → emulate viewport=2400x1260x2 → wait 2s → take_screenshot
+5. Downsample: magick {screenshot} -filter Lanczos -resize 2400x1260 -sampling-factor 4:4:4 -strip -quality 95 content/blog/<slug>/cover.png
+6. Set frontmatter: cover_image: cover.png, cover_image_alt: descriptive alt text
+
+## Dependencies
+### MCP
+- chrome-devtools — rendering
+
+### System
+- ImageMagick (magick CLI) — downsampling
+
+## Full Documentation
+.stitch/design.md (canonical 6-slot spec)
+docs/workflows/cover-images.md
diff --git a/.claude/skills/css-bem-migration/SKILL.md b/.claude/skills/css-bem-migration/SKILL.md
new file mode 100644
index 000000000..9fcd1046f
--- /dev/null
+++ b/.claude/skills/css-bem-migration/SKILL.md
@@ -0,0 +1,49 @@
+# CSS BEM Migration Skill
+#
+# Reusable methodology for migrating from FL-Builder (Beaver Builder)
+# auto-generated classes to BEM component architecture.
+
+## Overview
+TDD-driven FL-node extraction: add BEM classes → verify visual fidelity →
+remove FL-nodes → retest. Micro-commits on every green test.
+
+## Pattern
+1. RED: Identify FL-node CSS dependencies
+2. GREEN: Add BEM classes to HTML, consolidate CSS
+3. REFACTOR: Remove duplicate FL-node CSS, cleanup
+4. TEST: `env PRECOMPILED_ASSETS=false bin/rake test:critical`
+5. COMMIT: Micro-commit on green tests
+
+## BEM Component Structure
+```html
+<!-- Before (FL-node) -->
+<div class="fl-col fl-node-yx43bujcaiqn hover-box">
+
+<!-- Transition (BEM + FL retention) -->
+<div class="fl-col fl-col-small hover-box c-feature-card">
+
+<!-- Final (Pure BEM) -->
+<div class="c-feature-card c-feature-card--variant">
+```
+
+## CSS Consolidation Pattern
+```css
+/* Before: 3 separate FL-node selectors */
+.fl-node-yx43bujcaiqn { width: 32%; margin-right: 30px; }
+.fl-node-ktz4ipj39vd6 { width: 36%; margin: 0 30px; }
+.fl-node-m39uvorzy5g8 { width: 32%; margin-left: 30px; }
+
+/* After: DRY BEM pattern */
+.c-feature-card { width: 32%; margin-right: 30px; }
+.c-feature-card--second { width: 36%; margin: 0 30px; }
+.c-feature-card--third { margin-right: 0; }
+```
+
+## Constraints
+- Never modify vendor files or dynamic templates
+- Preserve cascade order: Critical → Layout → Component → Theme → Footer
+- Run BOTH bin/test (macOS) AND bin/dtest (Linux Docker) before commit
+
+## Full Documentation
+docs/workflows/css-consolidation.md
+docs/projects/2509-css-migration/css-loading-order-analysis.md
diff --git a/.claude/skills/visual-regression-testing/SKILL.md b/.claude/skills/visual-regression-testing/SKILL.md
new file mode 100644
index 000000000..ff12576f9
--- /dev/null
+++ b/.claude/skills/visual-regression-testing/SKILL.md
@@ -0,0 +1,41 @@
+# Visual Regression Testing Skill
+#
+# Reusable visual regression testing protocol for Hugo static sites.
+
+## Overview
+Screenshot comparison-based visual testing using capybara-screenshot-diff.
+Blocks commits on any visual regression during refactoring.
+
+## Protocol
+1. Capture baseline screenshots BEFORE changes
+2. Make CSS/HTML changes
+3. Capture comparison screenshots AFTER changes
+4. Compare with tolerance:
+   - 0.0 for refactoring (zero visual changes)
+   - ≤0.03 for new features only
+5. Block commit on any difference > tolerance
+
+## Test Pattern
+```ruby
+def test_hero_section_visual_consistency
+  visit "/services/fractional-cto/"
+  screenshot_path = take_screenshot("fractional_cto_hero")
+  assert_visual_match(screenshot_path, "baseline/fractional_cto_hero.png",
+                      tolerance: 0.03)
+end
+```
+
+## Prohibited Patterns (ZERO TOLERANCE)
+- CSS class existence testing
+- HTML structure counting
+- Framework-generated selector usage
+- Method existence testing
+
+## Commands
+- `bin/test` — visual regression (macOS)
+- `bin/dtest` — visual regression (Linux/Docker, CI parity)
+- BOTH must pass before commit for HTML/CSS changes
+
+## Full Documentation
+.apm/instructions/tdd-enforcement.instructions.md
+docs/60.03-tdd-quality-enforcement.md
diff --git a/.codex/agents/analyst.toml b/.codex/agents/analyst.toml
new file mode 100644
index 000000000..bdee948dd
--- /dev/null
+++ b/.codex/agents/analyst.toml
@@ -0,0 +1,3 @@
+name = "analyst"
+description = ""
+developer_instructions = "# analyst.agent.md — Thin Wrapper\n#\n# Full agent definitions: .claude/agents/core/analyst.md\n\n## Role\nAnalysis agent. Performs research, root-cause analysis, and pattern discovery.\n\n## Instructions\n- @.apm/instructions/project-rules.instructions.md\n- @.apm/prompts/research.prompt.md\n\n## Research Protocol (MANDATORY)\n1. claude-context search FIRST\n2. Verify with rg/ls\n3. Read docs/workflows/BASE_HANDBOOK.md"
diff --git a/.codex/agents/build-monitor.toml b/.codex/agents/build-monitor.toml
new file mode 100644
index 000000000..aea366de9
--- /dev/null
+++ b/.codex/agents/build-monitor.toml
@@ -0,0 +1,3 @@
+name = "build-monitor"
+description = ""
+developer_instructions = "# build-monitor.agent.md — Thin Wrapper\n#\n# Full agent definitions: .claude/agents/build-monitor.md\n\n## Role\nBuild monitoring agent. Validates Hugo builds, CI/CD pipelines, and performance.\n\n## Instructions\n- @.apm/instructions/project-rules.instructions.md\n\n## Gates\n- `bin/hugo-build` must pass (zero errors)\n- Chrome DevTools: zero console errors, zero 404s\n- `bin/rake test:critical` green on every commit\n\n## CI/CD\n- GitHub Actions → GitHub Pages\n- Workflows: .github/workflows/publish.yml, test.yml, sync-and-publish.yml"
diff --git a/.codex/agents/coder.toml b/.codex/agents/coder.toml
new file mode 100644
index 000000000..4ce6b3ebf
--- /dev/null
+++ b/.codex/agents/coder.toml
@@ -0,0 +1,3 @@
+name = "coder"
+description = ""
+developer_instructions = "# coder.agent.md — Thin Wrapper\n#\n# Canonical instructions: .apm/instructions/project-rules.instructions.md\n# Full agent definitions: .claude/agents/core/coder.md\n# \n# APM compiles this agent for all target tools. The canonical behavior\n# lives in the instruction files; this file pins the agent's role + pointers.\n\n## Role\nPrimary implementation agent. Writes code following TDD cycle (RED → GREEN → REFACTOR).\n\n## Instructions\n- @.apm/instructions/project-rules.instructions.md\n- @.apm/instructions/tdd-enforcement.instructions.md\n\n## For CSS/HTML work\n- @.apm/instructions/css-consolidation.instructions.md\n\n## For Ruby work\n- @docs/workflows/ruby.md\n\n## Constraints\n- Behavior-focused tests ONLY — reject structural/CSS class tests\n- Run `bin/rake test:critical` after every micro-change\n- Never use `npm` — use `bun`\n- Never create ad hoc `*.sh` test scripts"
diff --git a/.codex/agents/content-creator.toml b/.codex/agents/content-creator.toml
new file mode 100644
index 000000000..e9cb82791
--- /dev/null
+++ b/.codex/agents/content-creator.toml
@@ -0,0 +1,3 @@
+name = "content-creator"
+description = ""
+developer_instructions = "# content-creator.agent.md — Thin Wrapper\n#\n# Canonical instructions: .apm/instructions/voice-guide.instructions.md\n# Full agent definitions: .claude/agents/content-creator.md\n\n## Role\nContent creation agent. Writes blog posts, LinkedIn posts, and marketing copy.\n\n## Instructions\n- @.apm/instructions/voice-guide.instructions.md\n- @.apm/instructions/icp-target.instructions.md\n- @docs/workflows/blog-pipeline.md\n- @docs/workflows/linkedin-post-pipeline.md\n\n## Constraints\n- Zero generic AI language — run slop-detector + humanizer pass\n- Cold-eyes gate (9 checks) mandatory before publishing\n- Every sentence needs a person doing something\n- Banned words: unlock, harness, leverage, dive in, game-changer, etc.\n- Use `-` not `—` for all dashes"
diff --git a/.codex/agents/content-marketer.toml b/.codex/agents/content-marketer.toml
new file mode 100644
index 000000000..86d1e1cf5
--- /dev/null
+++ b/.codex/agents/content-marketer.toml
@@ -0,0 +1,3 @@
+name = "content-marketer"
+description = ""
+developer_instructions = "# content-marketer.agent.md — Thin Wrapper\n#\n# Full agent definitions: .claude/agents/content/content-marketer.md\n\n## Role\nContent marketing agent. Plans content calendars, researches topics, audits SEO performance.\n\n## Instructions\n- @.apm/instructions/icp-target.instructions.md\n- @docs/projects/2510-seo-content-strategy/20-29-strategy/20.07-content-plan-icp-e-q2-2026.md"
diff --git a/.codex/agents/hugo-expert.toml b/.codex/agents/hugo-expert.toml
new file mode 100644
index 000000000..e0413ba38
--- /dev/null
+++ b/.codex/agents/hugo-expert.toml
@@ -0,0 +1,3 @@
+name = "hugo-expert"
+description = ""
+developer_instructions = "# hugo-expert.agent.md — Thin Wrapper\n#\n# Full agent definitions: .claude/agents/hugo-expert.md\n\n## Role\nHugo static site generator expert. Handles templates, shortcodes, content structure, and build optimization.\n\n## Instructions\n- @.apm/instructions/project-rules.instructions.md\n- @docs/workflows/css-consolidation.md\n\n## Key Files\n- `config/_default/hugo.toml` — Hugo configuration\n- `themes/beaver/layouts/` — Templates\n- `themes/beaver/assets/css/` — CSS (PostCSS pipeline)\n- `content/` — Content files (Markdown + Hugo frontmatter)\n\n## Commands\n- `bin/hugo-build` — Build + validate\n- `hugo server -D` — Dev server\n\n## Constraints\n- Never modify vendor CSS files\n- Preserve cascade order: Critical → Layout → Component → Theme → Footer\n- BEM methodology for all new CSS"
diff --git a/.codex/agents/planner.toml b/.codex/agents/planner.toml
new file mode 100644
index 000000000..aae008394
--- /dev/null
+++ b/.codex/agents/planner.toml
@@ -0,0 +1,3 @@
+name = "planner"
+description = ""
+developer_instructions = "# planner.agent.md — Thin Wrapper\n#\n# Full agent definitions: .claude/agents/core/planner.md\n\n## Role\nPlanning agent. Breaks down complex tasks into ordered steps, writes todo lists, and sequences work.\n\n## Instructions\n- @.apm/instructions/project-rules.instructions.md"
diff --git a/.codex/agents/qa-browser-tester.toml b/.codex/agents/qa-browser-tester.toml
new file mode 100644
index 000000000..ed5c842a2
--- /dev/null
+++ b/.codex/agents/qa-browser-tester.toml
@@ -0,0 +1,3 @@
+name = "qa-browser-tester"
+description = ""
+developer_instructions = "# qa-browser-tester.agent.md — Thin Wrapper\n#\n# Full agent definitions: .claude/agents/validation/qa-browser-tester.md\n\n## Role\nQA browser tester. Runs Chrome DevTools validation after HTML/CSS/JS changes.\n\n## Instructions\n- @.apm/instructions/project-rules.instructions.md\n- @.apm/instructions/tdd-enforcement.instructions.md\n\n## Gates\n- Console: ZERO JS errors\n- Network: ZERO 404s for assets\n- Capture desktop + mobile screenshots\n- All gates pass → commit"
diff --git a/.codex/agents/researcher.toml b/.codex/agents/researcher.toml
new file mode 100644
index 000000000..facca9c49
--- /dev/null
+++ b/.codex/agents/researcher.toml
@@ -0,0 +1,3 @@
+name = "researcher"
+description = ""
+developer_instructions = "# researcher.agent.md — Thin Wrapper\n#\n# Full agent definitions: .claude/agents/core/researcher.md\n\n## Role\nResearch agent. Discovers existing patterns, framework docs, and best practices before implementation.\n\n## Instructions\n- @.apm/instructions/project-rules.instructions.md\n- @.apm/prompts/research.prompt.md"
diff --git a/.codex/agents/reviewer.toml b/.codex/agents/reviewer.toml
new file mode 100644
index 000000000..efe0f338d
--- /dev/null
+++ b/.codex/agents/reviewer.toml
@@ -0,0 +1,3 @@
+name = "reviewer"
+description = ""
+developer_instructions = "# reviewer.agent.md — Thin Wrapper\n#\n# Canonical instructions: .apm/instructions/project-rules.instructions.md\n# Full agent definitions: .claude/agents/core/reviewer.md\n\n## Role\nCode review agent. Validates correctness, adherence to project rules, and TDD compliance.\n\n## Instructions\n- @.apm/instructions/project-rules.instructions.md\n- @.apm/instructions/tdd-enforcement.instructions.md\n\n## Gates\n- Zero structural test patterns (CSS class testing, HTML counting)\n- Visual regression tolerance verified\n- Four-eyes principle enforced for all critical changes\n- Feature branch + PR pattern for HTML/CSS/template changes"
diff --git a/.codex/agents/ruby-coder.toml b/.codex/agents/ruby-coder.toml
new file mode 100644
index 000000000..b3791acf4
--- /dev/null
+++ b/.codex/agents/ruby-coder.toml
@@ -0,0 +1,3 @@
+name = "ruby-coder"
+description = ""
+developer_instructions = "# ruby-coder.agent.md — Thin Wrapper\n#\n# Full agent definitions: .claude/agents/ruby/ruby-coder.md\n\n## Role\nRuby implementation agent. Writes Ruby code (tests, sync scripts, lib/) following TDD.\n\n## Instructions\n- @.apm/instructions/project-rules.instructions.md\n- @.apm/instructions/tdd-enforcement.instructions.md\n- @docs/workflows/ruby.md\n\n## Commands\n- `bin/rake test:critical`\n- `bundle exec standardrb --fix`"
diff --git a/.codex/agents/ruby-reviewer.toml b/.codex/agents/ruby-reviewer.toml
new file mode 100644
index 000000000..4d9683359
--- /dev/null
+++ b/.codex/agents/ruby-reviewer.toml
@@ -0,0 +1,3 @@
+name = "ruby-reviewer"
+description = ""
+developer_instructions = "# ruby-reviewer.agent.md — Thin Wrapper\n#\n# Full agent definitions: .claude/agents/ruby/ruby-reviewer.md\n\n## Role\nRuby code reviewer. Validates Ruby code quality, TDD compliance, and Minitest patterns.\n\n## Instructions\n- @.apm/instructions/tdd-enforcement.instructions.md\n\n## Constraints\n- Behavior-focused tests ONLY — reject structural testing\n- StandardRB compliance\n- No ad hoc test scripts"
diff --git a/.codex/agents/screenshot-guardian.toml b/.codex/agents/screenshot-guardian.toml
new file mode 100644
index 000000000..7102da53f
--- /dev/null
+++ b/.codex/agents/screenshot-guardian.toml
@@ -0,0 +1,3 @@
+name = "screenshot-guardian"
+description = ""
+developer_instructions = "# screenshot-guardian.agent.md — Thin Wrapper\n#\n# Full agent definitions: .claude/agents/core/screenshot-guardian.md\n\n## Role\nVisual regression guardian. Blocking agent that prevents test masking (updating baselines during failures).\n\n## Instructions\n- @.apm/instructions/tdd-enforcement.instructions.md\n- @.claude/hooks/pre-commit-screenshot-validation.md\n\n## Gates\n- ZERO tolerance: screenshot updates during visual test failures\n- Tolerance: 0.0 for refactors, ≤0.03 for new features\n- Both bin/test AND bin/dtest must pass before commit"
diff --git a/.codex/agents/seo-specialist.toml b/.codex/agents/seo-specialist.toml
new file mode 100644
index 000000000..b49ca6465
--- /dev/null
+++ b/.codex/agents/seo-specialist.toml
@@ -0,0 +1,3 @@
+name = "seo-specialist"
+description = ""
+developer_instructions = "# seo-specialist.agent.md — Thin Wrapper\n#\n# Full agent definitions: .claude/agents/seo-specialist.md\n\n## Role\nSEO optimization agent. Handles keyword research, meta tags, schema, and content optimization.\n\n## Instructions\n- @docs/90-99-content-strategy/seo-optimization/\n- @docs/90-99-content-strategy/strategy-analysis/\n\n## Key Files\n- `themes/beaver/layouts/partials/seo/enhanced-meta-tags.html` — SEO/og:image template\n- `themes/beaver/layouts/partials/seo/article-schema.html` — JSON-LD schema\n- `config/_default/hugo.toml` — Hugo config (permalinks, sitemap)\n\n## Constraints\n- Title ≤45 chars, description 150-160 chars\n- Primary keyword in first 100 words\n- 4+ internal links, 5+ external citations\n- No keyword stuffing"
diff --git a/.codex/agents/tester.toml b/.codex/agents/tester.toml
new file mode 100644
index 000000000..39033b406
--- /dev/null
+++ b/.codex/agents/tester.toml
@@ -0,0 +1,3 @@
+name = "tester"
+description = ""
+developer_instructions = "# tester.agent.md — Thin Wrapper\n#\n# Canonical instructions: .apm/instructions/project-rules.instructions.md\n# Full agent definitions: .claude/agents/core/tester.md\n\n## Role\nTest agent. Writes and runs tests following TDD methodology.\n\n## Instructions\n- @.apm/instructions/project-rules.instructions.md\n- @.apm/instructions/tdd-enforcement.instructions.md\n\n## Framework\n- Minitest (test/system/, test/unit/)\n- capybara-screenshot-diff for visual regression\n\n## Commands\n- `bin/rake test:critical` — critical tests\n- `bin/test` — visual regression (macOS)\n- `bin/dtest` — visual regression (Linux/Docker)\n\n## Constraints\n- Behavior-focused tests ONLY\n- Assert shape, not config values\n- Run both bin/test AND bin/dtest before commit for HTML/CSS changes"
diff --git a/.codex/agents/tutorial-engineer.toml b/.codex/agents/tutorial-engineer.toml
new file mode 100644
index 000000000..0557214e1
--- /dev/null
+++ b/.codex/agents/tutorial-engineer.toml
@@ -0,0 +1,3 @@
+name = "tutorial-engineer"
+description = ""
+developer_instructions = "# tutorial-engineer.agent.md — Thin Wrapper\n#\n# Full agent definitions: .claude/agents/content/tutorial-engineer.md\n\n## Role\nTutorial engineer. Creates step-by-step how-to content, validates code examples, and ensures copy-paste-ability.\n\n## Instructions\n- @.apm/instructions/voice-guide.instructions.md\n- @.apm/instructions/icp-target.instructions.md"
diff --git a/.codex/agents/ui-problem-diagnosis.toml b/.codex/agents/ui-problem-diagnosis.toml
new file mode 100644
index 000000000..2e3643400
--- /dev/null
+++ b/.codex/agents/ui-problem-diagnosis.toml
@@ -0,0 +1,3 @@
+name = "ui-problem-diagnosis"
+description = ""
+developer_instructions = "# ui-problem-diagnosis.agent.md — Thin Wrapper\n#\n# Full agent definitions: .claude/agents/validation/ui-problem-diagnosis-specialist.md\n\n## Role\nUI problem diagnosis specialist. Investigates visual regressions, cross-browser issues, and CSS bugs.\n\n## Instructions\n- @.apm/instructions/css-consolidation.instructions.md\n- @docs/projects/2509-css-migration/css-loading-order-analysis.md"
diff --git a/.codex/agents/ux-browser-validator.toml b/.codex/agents/ux-browser-validator.toml
new file mode 100644
index 000000000..0a5eac942
--- /dev/null
+++ b/.codex/agents/ux-browser-validator.toml
@@ -0,0 +1,3 @@
+name = "ux-browser-validator"
+description = ""
+developer_instructions = "# ux-browser-validator.agent.md — Thin Wrapper\n#\n# Full agent definitions: .claude/agents/validation/ux-browser-validator.md\n\n## Role\nUX browser validator. Validates UX interactions, accessibility, and cross-browser behavior.\n\n## Instructions\n- @.apm/instructions/project-rules.instructions.md\n\n## Gates\n- WCAG 2.1 AA accessibility compliance\n- Semantic HTML elements (h1, button, main, nav)\n- Responsive design: desktop + mobile verification"
diff --git a/.codex/config.toml b/.codex/config.toml
new file mode 100644
index 000000000..be084c500
--- /dev/null
+++ b/.codex/config.toml
@@ -0,0 +1,50 @@
+[mcp_servers.claude-context]
+command = "npx"
+args = [ "-y", "@anthropic/claude-context",]
+id = ""
+
+[mcp_servers.context7]
+command = "npx"
+args = [ "-y", "@upstash/context7-mcp",]
+id = ""
+
+[mcp_servers.brave-search]
+command = "npx"
+args = [ "-y", "@anthropic/brave-search-mcp",]
+id = ""
+
+[mcp_servers.lean-ctx]
+command = "npx"
+args = [ "-y", "lean-ctx-mcp",]
+id = ""
+
+[mcp_servers.chrome-devtools]
+command = "npx"
+args = [ "-y", "@anthropic/chrome-devtools-mcp",]
+id = ""
+
+[mcp_servers.package-search]
+command = "npx"
+args = [ "-y", "package-search-mcp",]
+id = ""
+
+[mcp_servers.github-mcp-server]
+command = "npx"
+args = [ "-y", "@anthropic/github-mcp-server",]
+id = ""
+
+[mcp_servers.claude-context.env]
+
+[mcp_servers.context7.env]
+
+[mcp_servers.brave-search.env]
+BRAVE_API_KEY = "${BRAVE_API_KEY}"
+
+[mcp_servers.lean-ctx.env]
+
+[mcp_servers.chrome-devtools.env]
+
+[mcp_servers.package-search.env]
+
+[mcp_servers.github-mcp-server.env]
+GITHUB_TOKEN = "${GITHUB_TOKEN}"
diff --git a/.cursor/agents/analyst.md b/.cursor/agents/analyst.md
new file mode 100644
index 000000000..dfa4ca692
--- /dev/null
+++ b/.cursor/agents/analyst.md
@@ -0,0 +1,15 @@
+# analyst.agent.md — Thin Wrapper
+#
+# Full agent definitions: .claude/agents/core/analyst.md
+
+## Role
+Analysis agent. Performs research, root-cause analysis, and pattern discovery.
+
+## Instructions
+- @.apm/instructions/project-rules.instructions.md
+- @.apm/prompts/research.prompt.md
+
+## Research Protocol (MANDATORY)
+1. claude-context search FIRST
+2. Verify with rg/ls
+3. Read docs/workflows/BASE_HANDBOOK.md
diff --git a/.cursor/agents/build-monitor.md b/.cursor/agents/build-monitor.md
new file mode 100644
index 000000000..2bb9c0910
--- /dev/null
+++ b/.cursor/agents/build-monitor.md
@@ -0,0 +1,18 @@
+# build-monitor.agent.md — Thin Wrapper
+#
+# Full agent definitions: .claude/agents/build-monitor.md
+
+## Role
+Build monitoring agent. Validates Hugo builds, CI/CD pipelines, and performance.
+
+## Instructions
+- @.apm/instructions/project-rules.instructions.md
+
+## Gates
+- `bin/hugo-build` must pass (zero errors)
+- Chrome DevTools: zero console errors, zero 404s
+- `bin/rake test:critical` green on every commit
+
+## CI/CD
+- GitHub Actions → GitHub Pages
+- Workflows: .github/workflows/publish.yml, test.yml, sync-and-publish.yml
diff --git a/.cursor/agents/coder.md b/.cursor/agents/coder.md
new file mode 100644
index 000000000..7c8b2615e
--- /dev/null
+++ b/.cursor/agents/coder.md
@@ -0,0 +1,26 @@
+# coder.agent.md — Thin Wrapper
+#
+# Canonical instructions: .apm/instructions/project-rules.instructions.md
+# Full agent definitions: .claude/agents/core/coder.md
+# 
+# APM compiles this agent for all target tools. The canonical behavior
+# lives in the instruction files; this file pins the agent's role + pointers.
+
+## Role
+Primary implementation agent. Writes code following TDD cycle (RED → GREEN → REFACTOR).
+
+## Instructions
+- @.apm/instructions/project-rules.instructions.md
+- @.apm/instructions/tdd-enforcement.instructions.md
+
+## For CSS/HTML work
+- @.apm/instructions/css-consolidation.instructions.md
+
+## For Ruby work
+- @docs/workflows/ruby.md
+
+## Constraints
+- Behavior-focused tests ONLY — reject structural/CSS class tests
+- Run `bin/rake test:critical` after every micro-change
+- Never use `npm` — use `bun`
+- Never create ad hoc `*.sh` test scripts
diff --git a/.cursor/agents/content-creator.md b/.cursor/agents/content-creator.md
new file mode 100644
index 000000000..e6c244ca8
--- /dev/null
+++ b/.cursor/agents/content-creator.md
@@ -0,0 +1,20 @@
+# content-creator.agent.md — Thin Wrapper
+#
+# Canonical instructions: .apm/instructions/voice-guide.instructions.md
+# Full agent definitions: .claude/agents/content-creator.md
+
+## Role
+Content creation agent. Writes blog posts, LinkedIn posts, and marketing copy.
+
+## Instructions
+- @.apm/instructions/voice-guide.instructions.md
+- @.apm/instructions/icp-target.instructions.md
+- @docs/workflows/blog-pipeline.md
+- @docs/workflows/linkedin-post-pipeline.md
+
+## Constraints
+- Zero generic AI language — run slop-detector + humanizer pass
+- Cold-eyes gate (9 checks) mandatory before publishing
+- Every sentence needs a person doing something
+- Banned words: unlock, harness, leverage, dive in, game-changer, etc.
+- Use `-` not `—` for all dashes
diff --git a/.cursor/agents/content-marketer.md b/.cursor/agents/content-marketer.md
new file mode 100644
index 000000000..053ecd0a7
--- /dev/null
+++ b/.cursor/agents/content-marketer.md
@@ -0,0 +1,10 @@
+# content-marketer.agent.md — Thin Wrapper
+#
+# Full agent definitions: .claude/agents/content/content-marketer.md
+
+## Role
+Content marketing agent. Plans content calendars, researches topics, audits SEO performance.
+
+## Instructions
+- @.apm/instructions/icp-target.instructions.md
+- @docs/projects/2510-seo-content-strategy/20-29-strategy/20.07-content-plan-icp-e-q2-2026.md
diff --git a/.cursor/agents/hugo-expert.md b/.cursor/agents/hugo-expert.md
new file mode 100644
index 000000000..24fe61525
--- /dev/null
+++ b/.cursor/agents/hugo-expert.md
@@ -0,0 +1,25 @@
+# hugo-expert.agent.md — Thin Wrapper
+#
+# Full agent definitions: .claude/agents/hugo-expert.md
+
+## Role
+Hugo static site generator expert. Handles templates, shortcodes, content structure, and build optimization.
+
+## Instructions
+- @.apm/instructions/project-rules.instructions.md
+- @docs/workflows/css-consolidation.md
+
+## Key Files
+- `config/_default/hugo.toml` — Hugo configuration
+- `themes/beaver/layouts/` — Templates
+- `themes/beaver/assets/css/` — CSS (PostCSS pipeline)
+- `content/` — Content files (Markdown + Hugo frontmatter)
+
+## Commands
+- `bin/hugo-build` — Build + validate
+- `hugo server -D` — Dev server
+
+## Constraints
+- Never modify vendor CSS files
+- Preserve cascade order: Critical → Layout → Component → Theme → Footer
+- BEM methodology for all new CSS
diff --git a/.cursor/agents/planner.md b/.cursor/agents/planner.md
new file mode 100644
index 000000000..837cedb13
--- /dev/null
+++ b/.cursor/agents/planner.md
@@ -0,0 +1,9 @@
+# planner.agent.md — Thin Wrapper
+#
+# Full agent definitions: .claude/agents/core/planner.md
+
+## Role
+Planning agent. Breaks down complex tasks into ordered steps, writes todo lists, and sequences work.
+
+## Instructions
+- @.apm/instructions/project-rules.instructions.md
diff --git a/.cursor/agents/qa-browser-tester.md b/.cursor/agents/qa-browser-tester.md
new file mode 100644
index 000000000..3a18ca639
--- /dev/null
+++ b/.cursor/agents/qa-browser-tester.md
@@ -0,0 +1,16 @@
+# qa-browser-tester.agent.md — Thin Wrapper
+#
+# Full agent definitions: .claude/agents/validation/qa-browser-tester.md
+
+## Role
+QA browser tester. Runs Chrome DevTools validation after HTML/CSS/JS changes.
+
+## Instructions
+- @.apm/instructions/project-rules.instructions.md
+- @.apm/instructions/tdd-enforcement.instructions.md
+
+## Gates
+- Console: ZERO JS errors
+- Network: ZERO 404s for assets
+- Capture desktop + mobile screenshots
+- All gates pass → commit
diff --git a/.cursor/agents/researcher.md b/.cursor/agents/researcher.md
new file mode 100644
index 000000000..5d953465d
--- /dev/null
+++ b/.cursor/agents/researcher.md
@@ -0,0 +1,10 @@
+# researcher.agent.md — Thin Wrapper
+#
+# Full agent definitions: .claude/agents/core/researcher.md
+
+## Role
+Research agent. Discovers existing patterns, framework docs, and best practices before implementation.
+
+## Instructions
+- @.apm/instructions/project-rules.instructions.md
+- @.apm/prompts/research.prompt.md
diff --git a/.cursor/agents/reviewer.md b/.cursor/agents/reviewer.md
new file mode 100644
index 000000000..989624346
--- /dev/null
+++ b/.cursor/agents/reviewer.md
@@ -0,0 +1,17 @@
+# reviewer.agent.md — Thin Wrapper
+#
+# Canonical instructions: .apm/instructions/project-rules.instructions.md
+# Full agent definitions: .claude/agents/core/reviewer.md
+
+## Role
+Code review agent. Validates correctness, adherence to project rules, and TDD compliance.
+
+## Instructions
+- @.apm/instructions/project-rules.instructions.md
+- @.apm/instructions/tdd-enforcement.instructions.md
+
+## Gates
+- Zero structural test patterns (CSS class testing, HTML counting)
+- Visual regression tolerance verified
+- Four-eyes principle enforced for all critical changes
+- Feature branch + PR pattern for HTML/CSS/template changes
diff --git a/.cursor/agents/ruby-coder.md b/.cursor/agents/ruby-coder.md
new file mode 100644
index 000000000..510d186df
--- /dev/null
+++ b/.cursor/agents/ruby-coder.md
@@ -0,0 +1,15 @@
+# ruby-coder.agent.md — Thin Wrapper
+#
+# Full agent definitions: .claude/agents/ruby/ruby-coder.md
+
+## Role
+Ruby implementation agent. Writes Ruby code (tests, sync scripts, lib/) following TDD.
+
+## Instructions
+- @.apm/instructions/project-rules.instructions.md
+- @.apm/instructions/tdd-enforcement.instructions.md
+- @docs/workflows/ruby.md
+
+## Commands
+- `bin/rake test:critical`
+- `bundle exec standardrb --fix`
diff --git a/.cursor/agents/ruby-reviewer.md b/.cursor/agents/ruby-reviewer.md
new file mode 100644
index 000000000..bdfcf27b8
--- /dev/null
+++ b/.cursor/agents/ruby-reviewer.md
@@ -0,0 +1,14 @@
+# ruby-reviewer.agent.md — Thin Wrapper
+#
+# Full agent definitions: .claude/agents/ruby/ruby-reviewer.md
+
+## Role
+Ruby code reviewer. Validates Ruby code quality, TDD compliance, and Minitest patterns.
+
+## Instructions
+- @.apm/instructions/tdd-enforcement.instructions.md
+
+## Constraints
+- Behavior-focused tests ONLY — reject structural testing
+- StandardRB compliance
+- No ad hoc test scripts
diff --git a/.cursor/agents/screenshot-guardian.md b/.cursor/agents/screenshot-guardian.md
new file mode 100644
index 000000000..f160b3b75
--- /dev/null
+++ b/.cursor/agents/screenshot-guardian.md
@@ -0,0 +1,15 @@
+# screenshot-guardian.agent.md — Thin Wrapper
+#
+# Full agent definitions: .claude/agents/core/screenshot-guardian.md
+
+## Role
+Visual regression guardian. Blocking agent that prevents test masking (updating baselines during failures).
+
+## Instructions
+- @.apm/instructions/tdd-enforcement.instructions.md
+- @.claude/hooks/pre-commit-screenshot-validation.md
+
+## Gates
+- ZERO tolerance: screenshot updates during visual test failures
+- Tolerance: 0.0 for refactors, ≤0.03 for new features
+- Both bin/test AND bin/dtest must pass before commit
diff --git a/.cursor/agents/seo-specialist.md b/.cursor/agents/seo-specialist.md
new file mode 100644
index 000000000..328a80718
--- /dev/null
+++ b/.cursor/agents/seo-specialist.md
@@ -0,0 +1,21 @@
+# seo-specialist.agent.md — Thin Wrapper
+#
+# Full agent definitions: .claude/agents/seo-specialist.md
+
+## Role
+SEO optimization agent. Handles keyword research, meta tags, schema, and content optimization.
+
+## Instructions
+- @docs/90-99-content-strategy/seo-optimization/
+- @docs/90-99-content-strategy/strategy-analysis/
+
+## Key Files
+- `themes/beaver/layouts/partials/seo/enhanced-meta-tags.html` — SEO/og:image template
+- `themes/beaver/layouts/partials/seo/article-schema.html` — JSON-LD schema
+- `config/_default/hugo.toml` — Hugo config (permalinks, sitemap)
+
+## Constraints
+- Title ≤45 chars, description 150-160 chars
+- Primary keyword in first 100 words
+- 4+ internal links, 5+ external citations
+- No keyword stuffing
diff --git a/.cursor/agents/tester.md b/.cursor/agents/tester.md
new file mode 100644
index 000000000..8bc5f213b
--- /dev/null
+++ b/.cursor/agents/tester.md
@@ -0,0 +1,25 @@
+# tester.agent.md — Thin Wrapper
+#
+# Canonical instructions: .apm/instructions/project-rules.instructions.md
+# Full agent definitions: .claude/agents/core/tester.md
+
+## Role
+Test agent. Writes and runs tests following TDD methodology.
+
+## Instructions
+- @.apm/instructions/project-rules.instructions.md
+- @.apm/instructions/tdd-enforcement.instructions.md
+
+## Framework
+- Minitest (test/system/, test/unit/)
+- capybara-screenshot-diff for visual regression
+
+## Commands
+- `bin/rake test:critical` — critical tests
+- `bin/test` — visual regression (macOS)
+- `bin/dtest` — visual regression (Linux/Docker)
+
+## Constraints
+- Behavior-focused tests ONLY
+- Assert shape, not config values
+- Run both bin/test AND bin/dtest before commit for HTML/CSS changes
diff --git a/.cursor/agents/tutorial-engineer.md b/.cursor/agents/tutorial-engineer.md
new file mode 100644
index 000000000..14374cb5e
--- /dev/null
+++ b/.cursor/agents/tutorial-engineer.md
@@ -0,0 +1,10 @@
+# tutorial-engineer.agent.md — Thin Wrapper
+#
+# Full agent definitions: .claude/agents/content/tutorial-engineer.md
+
+## Role
+Tutorial engineer. Creates step-by-step how-to content, validates code examples, and ensures copy-paste-ability.
+
+## Instructions
+- @.apm/instructions/voice-guide.instructions.md
+- @.apm/instructions/icp-target.instructions.md
diff --git a/.cursor/agents/ui-problem-diagnosis.md b/.cursor/agents/ui-problem-diagnosis.md
new file mode 100644
index 000000000..9f88e69ee
--- /dev/null
+++ b/.cursor/agents/ui-problem-diagnosis.md
@@ -0,0 +1,10 @@
+# ui-problem-diagnosis.agent.md — Thin Wrapper
+#
+# Full agent definitions: .claude/agents/validation/ui-problem-diagnosis-specialist.md
+
+## Role
+UI problem diagnosis specialist. Investigates visual regressions, cross-browser issues, and CSS bugs.
+
+## Instructions
+- @.apm/instructions/css-consolidation.instructions.md
+- @docs/projects/2509-css-migration/css-loading-order-analysis.md
diff --git a/.cursor/agents/ux-browser-validator.md b/.cursor/agents/ux-browser-validator.md
new file mode 100644
index 000000000..118ea4ffa
--- /dev/null
+++ b/.cursor/agents/ux-browser-validator.md
@@ -0,0 +1,14 @@
+# ux-browser-validator.agent.md — Thin Wrapper
+#
+# Full agent definitions: .claude/agents/validation/ux-browser-validator.md
+
+## Role
+UX browser validator. Validates UX interactions, accessibility, and cross-browser behavior.
+
+## Instructions
+- @.apm/instructions/project-rules.instructions.md
+
+## Gates
+- WCAG 2.1 AA accessibility compliance
+- Semantic HTML elements (h1, button, main, nav)
+- Responsive design: desktop + mobile verification
diff --git a/.cursor/commands/blog-pipeline.md b/.cursor/commands/blog-pipeline.md
new file mode 100644
index 000000000..5c9dae536
--- /dev/null
+++ b/.cursor/commands/blog-pipeline.md
@@ -0,0 +1,79 @@
+---
+{}
+---
+
+# Blog Post Pipeline — Trigger Prompt
+#
+# Copy this prompt to write the next blog post from the content plan.
+# Full pipeline: docs/workflows/blog-pipeline.md
+#
+# This is a reusable PROMPT — APM compiles it to slash-commands for
+# each AI tool (e.g., /blog-post in Claude, #blog-post in Copilot).
+
+Write the next blog post following the full pipeline.
+
+## STEP 1 — Pick Topic
+Read the content plan: docs/projects/2510-seo-content-strategy/20-29-strategy/20.07-content-plan-icp-e-q2-2026.md
+Find the next 🔲 Planned topic. Confirm slug, primary keyword, target publish date.
+
+## STEP 2 — Read Context (mandatory, do not skip)
+- ICP: docs/90-99-content-strategy/strategy-analysis/90.10-icp-primary-website-target.md
+- Voice guide: docs/90-99-content-strategy/strategy-analysis/90.11-voice-guide.md
+- Content plan brief: read the topic's brief section
+- Design spec: .stitch/design.md
+
+## STEP 3 — Research
+- Web search for 5+ authoritative sources
+- Check existing posts for duplication: ls content/blog/ | grep -iE "<key-terms>"
+- Verify every statistic has a citable source
+
+## STEP 3b — Find Internal Posts to Reference
+- Use claude-context MCP search FIRST
+- For tag/slug lookups: docs/blog-post-index.md
+- Verify each slug: ls content/blog/<slug>/index.md
+- Collect at least 4 verified slugs before drafting
+
+## STEP 4 — Draft
+Create content/blog/<slug>/index.md with:
+- Frontmatter: title ≤45 chars, description 150-160 chars, date, slug, keywords, tags
+- Voice: plain English, empathy first, founder-to-founder tone
+- Structure: punchy hook → keyword in first 100 words → numbered H2s → "When NOT to" → JT proof → CTA → related reading → further reading
+- 4+ internal links in body, 5+ external citations
+- Paragraphs ≤3 sentences
+- Cut test: "what does the reader DO differently?" — delete if answer is "nothing new"
+
+## STEP 5 — Review Loop
+Spawn 3 parallel critic agents:
+1. Founder persona (Alex, burned CEO) — ≥8/10
+2. SEO/slop detector — ≥8/10 humanization
+3. Senior copy editor — "publish as-is" or "minor edits"
+
+## STEP 5c — Cold-Eyes Gate (mandatory)
+Spawn fresh-context reviewer with 9 explicit checks:
+1. Frontmatter vs body consistency
+2. Fluffy AI narration detection
+3. Fake authority stamps
+4. Timeline fabrication
+5. "Not X but Y" patterns
+6. Em dashes → hyphens
+7. Rule-of-three modifiers
+8. Cross-section template repetition
+9. Impersonal fragment stacking
+
+## STEP 6 — Cover Image
+- Read .stitch/design.md (6-slot layout)
+- Duplicate existing HTML from .stitch/designs/
+- Render: chrome-devtools → viewport 2400×1260×2 → screenshot
+- Downsample: magick → 2400×1260 → content/blog/<slug>/cover.png
+
+## STEP 7 — Validate
+- bin/hugo-build must pass (zero errors)
+- Chrome DevTools: zero console errors, zero 404s, og:image resolves
+
+## STEP 8 — Update Content Plan
+- Change topic status: 🔲 Planned → ✅ Published
+- Add commit SHA + changelog
+
+## STEP 9 — Commit
+- Stage: post directory + cover HTML + content plan update
+- Commit format: feat(content): add <topic> post
\ No newline at end of file
diff --git a/.cursor/commands/cover-image.md b/.cursor/commands/cover-image.md
new file mode 100644
index 000000000..024cd8573
--- /dev/null
+++ b/.cursor/commands/cover-image.md
@@ -0,0 +1,23 @@
+---
+{}
+---
+
+# Cover Image Generation — Trigger Prompt
+#
+# Reusable prompt for generating blog post cover images via Stitch design system.
+# Full spec: .stitch/design.md (canonical 6-slot layout)
+# Full workflow: docs/workflows/cover-images.md
+
+Generate a cover image for a blog post.
+
+## Steps
+1. Read .stitch/design.md for the 6-slot layout specification
+2. Duplicate an existing HTML from .stitch/designs/ (e.g., rails-8-1-active-job-continuations-cover.html)
+3. Change ONLY: title, category, pill, 3 headline lines, 3 chip labels+values, status text
+4. Render: chrome-devtools new_page → emulate viewport=2400x1260x2 → wait 2s → take_screenshot
+5. Downsample: magick {screenshot} -filter Lanczos -resize 2400x1260 -sampling-factor 4:4:4 -strip -quality 95 content/blog/<slug>/cover.png
+
+## Design System
+- JetVelocity: obsidian dark, Ruby red (#cc342d), neon purple (#a855f7)
+- Font: Inter (headlines), JetBrains Mono (code/metrics)
+- 6-slot grid layout — see .stitch/design.md for exact positioning
\ No newline at end of file
diff --git a/.cursor/commands/research.md b/.cursor/commands/research.md
new file mode 100644
index 000000000..59101394b
--- /dev/null
+++ b/.cursor/commands/research.md
@@ -0,0 +1,30 @@
+---
+{}
+---
+
+# Research Protocol — Trigger Prompt
+#
+# Reusable prompt for research-first development.
+# MANDATORY sequence before any implementation.
+
+Research this topic before implementing: {TOPIC}
+
+## Research Sequence
+1. **Handbook search**: Search claude-context for existing patterns in /knowledge/ and docs/
+2. **Codebase search**: Search claude-context for similar implementations in this repo
+3. **Framework docs**: context7 for official documentation (Hugo, PostCSS, Ruby, etc.)
+4. **Package analysis**: package-search for dependency research (if using new packages)
+5. **Web search**: brave-search / searxng for current best practices (2025-2026)
+
+## Output Requirements
+- 5+ authoritative sources
+- Existing patterns found (or confirmation of none)
+- Framework documentation citations
+- Dependency evaluation (if applicable)
+- Recommendation with trade-offs
+
+## Anti-Patterns (BLOCKED)
+- Implementation before research
+- Single-source decisions
+- Dependencies installed without package-search verification
+- Slugs guessed without ls content/blog/ verification
\ No newline at end of file
diff --git a/.cursor/mcp.json b/.cursor/mcp.json
new file mode 100644
index 000000000..7d37ea206
--- /dev/null
+++ b/.cursor/mcp.json
@@ -0,0 +1,78 @@
+{
+  "mcpServers": {
+    "claude-context": {
+      "type": "stdio",
+      "command": "npx",
+      "args": [
+        "-y",
+        "@anthropic/claude-context"
+      ]
+    },
+    "context7": {
+      "type": "stdio",
+      "command": "npx",
+      "args": [
+        "-y",
+        "@upstash/context7-mcp"
+      ]
+    },
+    "brave-search": {
+      "type": "stdio",
+      "command": "npx",
+      "env": {
+        "BRAVE_API_KEY": "${BRAVE_API_KEY}"
+      },
+      "args": [
+        "-y",
+        "@anthropic/brave-search-mcp"
+      ]
+    },
+    "lean-ctx": {
+      "type": "stdio",
+      "command": "npx",
+      "args": [
+        "-y",
+        "lean-ctx-mcp"
+      ]
+    },
+    "chrome-devtools": {
+      "type": "stdio",
+      "command": "npx",
+      "args": [
+        "-y",
+        "@anthropic/chrome-devtools-mcp"
+      ]
+    },
+    "package-search": {
+      "type": "stdio",
+      "command": "npx",
+      "args": [
+        "-y",
+        "package-search-mcp"
+      ]
+    },
+    "github-mcp-server": {
+      "type": "stdio",
+      "command": "npx",
+      "env": {
+        "GITHUB_TOKEN": "${GITHUB_TOKEN}"
+      },
+      "args": [
+        "-y",
+        "@anthropic/github-mcp-server"
+      ]
+    },
+    "searxng": {
+      "type": "http",
+      "url": "http://localhost:8080"
+    },
+    "ollama": {
+      "type": "http",
+      "url": "http://localhost:11434"
+    },
+    "milvus": {
+      "type": "http",
+      "url": "http://localhost:19530"
+    }
+  }
+}
\ No newline at end of file
diff --git a/.cursor/rules/css-consolidation.mdc b/.cursor/rules/css-consolidation.mdc
new file mode 100644
index 000000000..c011f6ced
--- /dev/null
+++ b/.cursor/rules/css-consolidation.mdc
@@ -0,0 +1,34 @@
+---
+description: CSS Consolidation Rules — JetThoughts Site
+---
+
+# CSS Consolidation Rules — JetThoughts Site
+#
+# MANDATORY for all CSS/HTML template changes.
+# Full docs: docs/workflows/css-consolidation.md
+# Project: docs/projects/2509-css-migration/css-loading-order-analysis.md
+
+## Architecture
+- **Methodology**: BEM (Block__Element--Modifier)
+- **Pipeline**: PostCSS (import, mixins, nested, autoprefixer, cssnano, purgecss)
+- **Cascade order**: Critical → Layout → Component → Theme → Footer
+
+## Rules
+- Never modify vendor files (`css/vendors/*.min.css`) or dynamic templates (`css/dynamic-*.css`)
+- Preserve cascade order in all changes
+- Extract whole rule sets; keep page-specific `.fl-node-{hash}` in original files
+- Run `bin/rake test:critical` after each extraction
+
+## FL-Node Extraction (Beaver Builder → BEM)
+- **Phase**: TDD RED-GREEN-REFACTOR cycle
+- **Validation**: `env PRECOMPILED_ASSETS=false bin/rake test:critical` after each change
+- **Pattern**: Add BEM classes first → verify visual fidelity → remove FL-nodes → retest
+- **Commit**: micro-commit on green tests only
+
+## Visual Regression
+- Tolerance: 0.0 for refactoring (zero visual changes)
+- Tolerance: ≤0.03 for new features only
+- Run BOTH `bin/test` and `bin/dtest` before commit
+- Visual failures are commit blockers, not warnings
+
+@see docs/workflows/css-consolidation.md
diff --git a/.cursor/rules/icp-target.mdc b/.cursor/rules/icp-target.mdc
new file mode 100644
index 000000000..4c8fe7487
--- /dev/null
+++ b/.cursor/rules/icp-target.mdc
@@ -0,0 +1,50 @@
+---
+description: ICP Target — JetThoughts Primary Website Audience
+---
+
+# ICP Target — JetThoughts Primary Website Audience
+#
+# CANONICAL SOURCE: docs/90-99-content-strategy/strategy-analysis/90.10-icp-primary-website-target.md
+# This file is a COMPILED EXTRACT for APM — the full doc is authoritative.
+# Update the canonical source first, then sync this extract.
+#
+# MANDATORY reading before content, design, marketing, or blog work.
+# Compiled by APM into every AI tool's context.
+
+## ICP-E: Non-Technical Founder Burned by a Devshop
+
+### Who They Are
+- 35-50 years old
+- Non-technical or semi-technical
+- Raised funding ($500K–$2M seed)
+- Hired a dev shop or freelancers
+- Missed deadlines, buggy product, mounting technical debt
+- Starting to question: "Is it me? Is it them? How do I know?"
+
+### Their Pain
+- Can't evaluate code quality themselves
+- Dev shop says "almost done" for months
+- Every new feature breaks something else
+- No tests, no documentation, no deploy pipeline
+- Starting to lose investor confidence
+
+### What They Need
+- Someone who speaks plain English, not jargon
+- Proof that the problem is real and solvable (specific stories, not vague promises)
+- A clear path out: audit → stabilize → build right
+- Trust signals: named clients, specific results, honest trade-offs
+
+### What Resonates
+- Stories of other founders in the same situation
+- Specific numbers: "$40K wasted," "11-minute outage," "84 support tickets"
+- Honest acknowledgment that rescue is hard and expensive
+- The idea that good engineering is a competitive advantage
+- "You're not crazy — this IS what bad engineering looks like"
+
+### What Alienates
+- jargon-dense technical explanations before establishing trust
+- "We're the best" claims without evidence
+- Generic consulting language ("partner with you," "drive outcomes")
+- Ignoring the emotional reality (shame, frustration, fear of looking stupid)
+
+@see docs/90-99-content-strategy/strategy-analysis/90.10-icp-primary-website-target.md
diff --git a/.cursor/rules/project-rules.mdc b/.cursor/rules/project-rules.mdc
new file mode 100644
index 000000000..bf8184fd1
--- /dev/null
+++ b/.cursor/rules/project-rules.mdc
@@ -0,0 +1,83 @@
+---
+description: JetThoughts Site — Project Rules
+---
+
+# JetThoughts Site — Project Rules
+#
+# This is the SINGLE SOURCE for AI agent instructions across all tools.
+# APM compiles this to each tool's expected path (CLAUDE.md, .github/
+# copilot-instructions.md, etc.). Edit here, not in tool-specific files.
+#
+# CANONICAL SOURCE: THIS FILE. The existing CLAUDE.md is a Claude-specific
+# superset; over time CLI tool instructions converge here. Voice, ICP, TDD,
+# and CSS rules are split into separate instruction files for modularity.
+#
+# Full workflow docs: docs/workflows/
+# Full content docs: docs/90-99-content-strategy/
+
+## Stack
+- **Site**: Hugo 0.147.9 Extended static site generator
+- **Theme**: Custom "Beaver" theme (BEM + PostCSS)
+- **Tests**: Ruby 3.4, Minitest, Capybara, Selenium WebDriver
+- **CSS**: PostCSS pipeline (autoprefixer, cssnano, purgecss)
+- **Package managers**: Bun (JS, NOT npm), Bundler (Ruby)
+- **CI/CD**: GitHub Actions → GitHub Pages
+
+## Build & Test Commands
+- Build: `bin/hugo-build`
+- Test (critical): `bin/rake test:critical`
+- Test (visual/macOS): `bin/test`
+- Test (visual/Linux Docker): `bin/dtest`
+- Lint Ruby: `bundle exec standardrb --fix`
+
+## Research Protocol (MANDATORY — before any change)
+1. Search codebase with `claude-context` MCP FIRST
+2. Verify exact filenames with `rg`/`ls` as fallback
+3. Read `docs/workflows/BASE_HANDBOOK.md` at session start
+4. Read `docs/workflows/flow-router.md` to route to correct workflow
+5. For blog content: find existing posts via `docs/blog-post-index.md`
+6. Never guess slugs — verify with `ls content/blog/<slug>/index.md`
+
+## Non-Negotiables
+- **Zero generic AI language** in content — all AI-sounding phrases rejected
+- **Zero Hugo build breaks** — `bin/hugo-build` must pass for content changes
+- **Zero unsupported claims** — all assertions must have citations
+- **No ad hoc test scripts** — use `bin/rake test:critical` and `bin/test`
+- **No Python for analysis** — prefer `rg`, `sed`, `awk`, shell tools
+- **No duplicate throwaway files** — no `*_refactored.*`, `*_new.*`, `*_backup.*`
+- **New docs only under `docs/workflows/`** — else edit existing files
+- **Never commit coordinator/verification reports** — write to `/tmp/` instead
+
+## Content Work (Blog Posts, LinkedIn, Pages)
+- **BEFORE writing**: Read voice guide + ICP target (see instructions/voice-guide, instructions/icp-target)
+- **Blog posts**: Follow `docs/workflows/blog-pipeline.md` end-to-end
+- **LinkedIn**: Follow `docs/workflows/linkedin-post-pipeline.md`
+- **Cover images**: Follow `docs/workflows/cover-images.md` + `.stitch/design.md`
+- Run slop-detector + humanizer pass on all drafts
+- Cold-eyes gate (9 checks) mandatory before publishing
+
+## CSS / HTML Changes
+- **BEM methodology** mandatory for all new CSS
+- **FL-node extraction**: TDD cycle — RED → GREEN → REFACTOR → TEST
+- **Visual regression tolerance**: 0.0 for refactors, ≤0.03 for new features
+- **MANDATORY**: Run BOTH `bin/test` (macOS) AND `bin/dtest` (Linux Docker) before commit
+- **Feature branches**: HTML/CSS/template changes → feature branch → one PR per sprint
+- Zero tolerance: CSS class existence testing, framework selector usage in tests
+
+## TDD & Testing
+- **Cycle**: RED → GREEN → REFACTOR
+- **Framework**: Minitest (test/system/, test/unit/)
+- **Behavior-focused tests ONLY** — reject implementation/existence/config tests
+- **Avoid fragile config assertions** — assert shape, not tunable values
+- **Visual regression**: screenshot comparison, 0.0 tolerance for refactoring
+- Run `bin/rake test:critical` after every micro-change (<10 lines)
+- Chrome DevTools validation after HTML/CSS/JS changes: zero console errors, zero 404s
+
+## Behavioral Constraints
+- **User says "code is bad" or "over-engineered"**: HALT, 5-Why analysis, fix config, proceed
+- **Reflection triggers**: zero-defect tracking, test masking detection
+- **Four-eyes principle**: paired agents for all critical changes
+- **Direct-to-master only for**: content-only blog edits, commit fixes, docs, CLAUDE.md policy updates
+
+@see docs/workflows/ for full workflow documentation
+@see docs/90-99-content-strategy/ for content strategy
diff --git a/.cursor/rules/tdd-enforcement.mdc b/.cursor/rules/tdd-enforcement.mdc
new file mode 100644
index 000000000..941d9f481
--- /dev/null
+++ b/.cursor/rules/tdd-enforcement.mdc
@@ -0,0 +1,66 @@
+---
+description: TDD Enforcement — JetThoughts Site
+---
+
+# TDD Enforcement — JetThoughts Site
+#
+# MANDATORY for all code changes. Extracted from CLAUDE.md TDD section.
+# Full docs: docs/60.03-tdd-quality-enforcement.md, docs/60.04-four-eyes-principle.md
+
+## TDD Cycle
+1. **RED**: Write a failing test that validates BEHAVIOR (not implementation)
+2. **GREEN**: Minimal code to make the test pass
+3. **REFACTOR**: Clean up while tests stay green
+4. **COMMIT**: Micro-commit on green tests
+
+## Test Quality Rules (ZERO TOLERANCE)
+
+### Prohibited Test Patterns — Agent MUST Reject
+- ❌ CSS class existence testing (`.fl-node-*`, `.has_css?(".px-4")`)
+- ❌ HTML structure counting (`page.all("div").count`)
+- ❌ Framework-generated selector usage (FL-Builder, Tailwind utility classes)
+- ❌ Method existence testing (`assert SomeClass.method_defined?`)
+- ❌ Config value assertions (hardcoded `q=90`, exact file sizes, specific dimensions)
+- ❌ Implementation detail testing
+
+### Required Test Patterns
+- ✅ Visual regression: `take_screenshot` → `assert_visual_match` with tolerance
+- ✅ User-visible content: `assert page.has_content?("text")`
+- ✅ Semantic elements: `assert page.has_css?("h1", text: "...")`
+- ✅ User interaction: `click_button` → `assert_current_path`
+- ✅ Shape assertions: `q=\d+`, `src contains wsrv.nl`, not `q=90`, not exact URLs
+
+### Avoid Fragile Config Assertions
+Don't hardcode tunable values. Assert the *shape*, not the configuration:
+- Bad: `assert_match /q=90/, src`
+- Good: `assert_match /q=\d+/, src`
+- Bad: `assert_match /w=360/, src`
+- Good: `assert has_css?("picture source")`
+
+## Visual Regression Gates (BLOCKING for HTML/CSS changes)
+- **Tolerance**: 0.0 for refactoring, ≤0.03 for new features
+- **Protocol**: Capture baseline BEFORE changes → compare AFTER → block on any diff >0%
+- **MANDATORY**: Run BOTH `bin/test` (macOS) AND `bin/dtest` (Linux Docker) before commit
+- Visual regression failures are COMMIT BLOCKERS, not warnings
+
+## Test Runner
+- Framework: Minitest
+- Critical tests: `bin/rake test:critical` (after every micro-change <10 lines)
+- Visual macOS: `bin/test`
+- Visual Linux/Docker: `bin/dtest`
+- NEVER create ad hoc `*.sh` test scripts
+
+## Chrome DevTools Validation (MANDATORY after HTML/CSS/JS)
+1. Start Hugo dev server, open page in Chrome DevTools
+2. Console: ZERO JS errors
+3. Network: ZERO 404s
+4. Capture desktop + mobile screenshots
+5. All gates pass → commit. Any gate fails → fix, repeat
+
+## Feature Branch + PR (HTML/CSS/template changes)
+1. `git checkout -b <sprint-name>`
+2. Multiple related commits on branch
+3. Both test gates green on each commit
+4. `git push -u origin <branch>`
+5. ONE PR per sprint via `gh pr create`
+6. Direct-to-master only for: content-only blog edits, commit fixes, docs, CLAUDE.md updates
diff --git a/.cursor/rules/voice-guide.mdc b/.cursor/rules/voice-guide.mdc
new file mode 100644
index 000000000..5ef863934
--- /dev/null
+++ b/.cursor/rules/voice-guide.mdc
@@ -0,0 +1,72 @@
+---
+description: Voice Guide — JetThoughts Content
+---
+
+# Voice Guide — JetThoughts Content
+#
+# CANONICAL SOURCE: docs/90-99-content-strategy/strategy-analysis/90.11-voice-guide.md
+# This file is a COMPILED EXTRACT for APM — the full doc is authoritative.
+# Update the canonical source first, then sync this extract.
+#
+# MANDATORY reading before writing ANY blog post, LinkedIn post, or marketing copy.
+# Compiled by APM into every AI tool's context.
+
+## Two Voices
+
+### Founder Posts (Paul Keen / JT leadership)
+- **Rob Walling directness** + **Rand Fishkin vulnerability**
+- Plain English, empathy, specific numbers
+- Talk to a non-technical founder burned by a devshop
+- "Been there" energy — practitioner not vendor
+- Trade-offs acknowledged honestly
+
+### Tech Posts (engineering team)
+- **Thoughtbot style**: open with tension, not features
+- Own your opinions — no "it depends" hedging
+- Code as narrative evidence, not decoration
+- Honest trade-offs: what does this solution fail to do?
+
+## Banned Words
+unlock, harness, leverage, dive in, game-changer, journey, seamless, robust,
+supercharge, revolutionize, embark, delve, discover, blast radius, suspect list
+
+## Banned Structural Patterns
+- Rule of three (parallel triads)
+- Signposting ("This post is about...", "Here's what you need to know")
+- Bold inline-header lists
+- Triple rhetorical questions
+- Slogany parallel closings ("X is A. Y is B.")
+- Therapist voice ("That's the hardest part")
+- Copula avoidance ("serves as" → use "is")
+- Impersonal fragment stacking — every sentence needs a human subject
+- Sustained staccato (3+ consecutive short fragments or uniform-length sentences)
+- Command structure repetition (3+ paragraphs starting with imperatives)
+- Apologetic caveats ("every project is different")
+- Telling instead of showing — replace adjectives with concrete mechanics
+- Fake authority generalizations ("We've seen this on every codebase")
+- Timeline fabrication — verify technology release dates before using "last year"
+- Use `-` not `—` for all dashes
+
+## Quality Tests (run on every paragraph)
+- **"Who" test**: who is doing what? If subject is a thing, rewrite with a person
+- **"Show" test**: did I show the mechanic or just label it?
+- **"Coffee" test**: would a tired founder talk like this at a bar?
+- **"90/10" test**: is ≥90% educational? Is promotion ≤10%?
+- **"Trade-off" test**: did I acknowledge what the solution fails to do?
+- **"Practitioner" test**: replace generalized scenarios with specific incidents
+
+## Mannerism Noun Cap
+discipline, cycle, pattern, canonical, workflow, framework, cadence, rhythm,
+contract, gate, gatekeeper, layer, round, review, pass
+→ Max 5 occurrences total per post. Vary to the concrete thing happening.
+
+## Cadence Quotas (per H2 section)
+- At least one sentence <9 words AND one >22 words
+- No 3 consecutive sentences within 3 words of each other
+- One single-sentence paragraph per H2
+- Max 3 sentences / 70 words per paragraph
+- No two sentences in same paragraph starting with same first word
+- "The..." paragraph openers: <20% of total paragraphs
+- "We..." paragraph openers: <15% of total paragraphs
+
+@see docs/90-99-content-strategy/strategy-analysis/90.11-voice-guide.md
diff --git a/.gemini/commands/blog-pipeline.toml b/.gemini/commands/blog-pipeline.toml
new file mode 100644
index 000000000..60750cf6c
--- /dev/null
+++ b/.gemini/commands/blog-pipeline.toml
@@ -0,0 +1 @@
+prompt = "# Blog Post Pipeline — Trigger Prompt\n#\n# Copy this prompt to write the next blog post from the content plan.\n# Full pipeline: docs/workflows/blog-pipeline.md\n#\n# This is a reusable PROMPT — APM compiles it to slash-commands for\n# each AI tool (e.g., /blog-post in Claude, #blog-post in Copilot).\n\nWrite the next blog post following the full pipeline.\n\n## STEP 1 — Pick Topic\nRead the content plan: docs/projects/2510-seo-content-strategy/20-29-strategy/20.07-content-plan-icp-e-q2-2026.md\nFind the next 🔲 Planned topic. Confirm slug, primary keyword, target publish date.\n\n## STEP 2 — Read Context (mandatory, do not skip)\n- ICP: docs/90-99-content-strategy/strategy-analysis/90.10-icp-primary-website-target.md\n- Voice guide: docs/90-99-content-strategy/strategy-analysis/90.11-voice-guide.md\n- Content plan brief: read the topic's brief section\n- Design spec: .stitch/design.md\n\n## STEP 3 — Research\n- Web search for 5+ authoritative sources\n- Check existing posts for duplication: ls content/blog/ | grep -iE \"<key-terms>\"\n- Verify every statistic has a citable source\n\n## STEP 3b — Find Internal Posts to Reference\n- Use claude-context MCP search FIRST\n- For tag/slug lookups: docs/blog-post-index.md\n- Verify each slug: ls content/blog/<slug>/index.md\n- Collect at least 4 verified slugs before drafting\n\n## STEP 4 — Draft\nCreate content/blog/<slug>/index.md with:\n- Frontmatter: title ≤45 chars, description 150-160 chars, date, slug, keywords, tags\n- Voice: plain English, empathy first, founder-to-founder tone\n- Structure: punchy hook → keyword in first 100 words → numbered H2s → \"When NOT to\" → JT proof → CTA → related reading → further reading\n- 4+ internal links in body, 5+ external citations\n- Paragraphs ≤3 sentences\n- Cut test: \"what does the reader DO differently?\" — delete if answer is \"nothing new\"\n\n## STEP 5 — Review Loop\nSpawn 3 parallel critic agents:\n1. Founder persona (Alex, burned CEO) — ≥8/10\n2. SEO/slop detector — ≥8/10 humanization\n3. Senior copy editor — \"publish as-is\" or \"minor edits\"\n\n## STEP 5c — Cold-Eyes Gate (mandatory)\nSpawn fresh-context reviewer with 9 explicit checks:\n1. Frontmatter vs body consistency\n2. Fluffy AI narration detection\n3. Fake authority stamps\n4. Timeline fabrication\n5. \"Not X but Y\" patterns\n6. Em dashes → hyphens\n7. Rule-of-three modifiers\n8. Cross-section template repetition\n9. Impersonal fragment stacking\n\n## STEP 6 — Cover Image\n- Read .stitch/design.md (6-slot layout)\n- Duplicate existing HTML from .stitch/designs/\n- Render: chrome-devtools → viewport 2400×1260×2 → screenshot\n- Downsample: magick → 2400×1260 → content/blog/<slug>/cover.png\n\n## STEP 7 — Validate\n- bin/hugo-build must pass (zero errors)\n- Chrome DevTools: zero console errors, zero 404s, og:image resolves\n\n## STEP 8 — Update Content Plan\n- Change topic status: 🔲 Planned → ✅ Published\n- Add commit SHA + changelog\n\n## STEP 9 — Commit\n- Stage: post directory + cover HTML + content plan update\n- Commit format: feat(content): add <topic> post"
diff --git a/.gemini/commands/cover-image.toml b/.gemini/commands/cover-image.toml
new file mode 100644
index 000000000..54ad707dd
--- /dev/null
+++ b/.gemini/commands/cover-image.toml
@@ -0,0 +1 @@
+prompt = "# Cover Image Generation — Trigger Prompt\n#\n# Reusable prompt for generating blog post cover images via Stitch design system.\n# Full spec: .stitch/design.md (canonical 6-slot layout)\n# Full workflow: docs/workflows/cover-images.md\n\nGenerate a cover image for a blog post.\n\n## Steps\n1. Read .stitch/design.md for the 6-slot layout specification\n2. Duplicate an existing HTML from .stitch/designs/ (e.g., rails-8-1-active-job-continuations-cover.html)\n3. Change ONLY: title, category, pill, 3 headline lines, 3 chip labels+values, status text\n4. Render: chrome-devtools new_page → emulate viewport=2400x1260x2 → wait 2s → take_screenshot\n5. Downsample: magick {screenshot} -filter Lanczos -resize 2400x1260 -sampling-factor 4:4:4 -strip -quality 95 content/blog/<slug>/cover.png\n\n## Design System\n- JetVelocity: obsidian dark, Ruby red (#cc342d), neon purple (#a855f7)\n- Font: Inter (headlines), JetBrains Mono (code/metrics)\n- 6-slot grid layout — see .stitch/design.md for exact positioning"
diff --git a/.gemini/commands/research.toml b/.gemini/commands/research.toml
new file mode 100644
index 000000000..0947dd016
--- /dev/null
+++ b/.gemini/commands/research.toml
@@ -0,0 +1 @@
+prompt = "# Research Protocol — Trigger Prompt\n#\n# Reusable prompt for research-first development.\n# MANDATORY sequence before any implementation.\n\nResearch this topic before implementing: {TOPIC}\n\n## Research Sequence\n1. **Handbook search**: Search claude-context for existing patterns in /knowledge/ and docs/\n2. **Codebase search**: Search claude-context for similar implementations in this repo\n3. **Framework docs**: context7 for official documentation (Hugo, PostCSS, Ruby, etc.)\n4. **Package analysis**: package-search for dependency research (if using new packages)\n5. **Web search**: brave-search / searxng for current best practices (2025-2026)\n\n## Output Requirements\n- 5+ authoritative sources\n- Existing patterns found (or confirmation of none)\n- Framework documentation citations\n- Dependency evaluation (if applicable)\n- Recommendation with trade-offs\n\n## Anti-Patterns (BLOCKED)\n- Implementation before research\n- Single-source decisions\n- Dependencies installed without package-search verification\n- Slugs guessed without ls content/blog/ verification"
diff --git a/.gemini/settings.json b/.gemini/settings.json
new file mode 100644
index 000000000..46037b8c2
--- /dev/null
+++ b/.gemini/settings.json
@@ -0,0 +1,68 @@
+{
+  "mcpServers": {
+    "claude-context": {
+      "command": "npx",
+      "args": [
+        "-y",
+        "@anthropic/claude-context"
+      ]
+    },
+    "context7": {
+      "command": "npx",
+      "args": [
+        "-y",
+        "@upstash/context7-mcp"
+      ]
+    },
+    "brave-search": {
+      "command": "npx",
+      "env": {
+        "BRAVE_API_KEY": "${BRAVE_API_KEY}"
+      },
+      "args": [
+        "-y",
+        "@anthropic/brave-search-mcp"
+      ]
+    },
+    "lean-ctx": {
+      "command": "npx",
+      "args": [
+        "-y",
+        "lean-ctx-mcp"
+      ]
+    },
+    "chrome-devtools": {
+      "command": "npx",
+      "args": [
+        "-y",
+        "@anthropic/chrome-devtools-mcp"
+      ]
+    },
+    "package-search": {
+      "command": "npx",
+      "args": [
+        "-y",
+        "package-search-mcp"
+      ]
+    },
+    "github-mcp-server": {
+      "command": "npx",
+      "env": {
+        "GITHUB_TOKEN": "${GITHUB_TOKEN}"
+      },
+      "args": [
+        "-y",
+        "@anthropic/github-mcp-server"
+      ]
+    },
+    "searxng": {
+      "httpUrl": "http://localhost:8080"
+    },
+    "ollama": {
+      "httpUrl": "http://localhost:11434"
+    },
+    "milvus": {
+      "httpUrl": "http://localhost:19530"
+    }
+  }
+}
\ No newline at end of file
diff --git a/.github/agents/analyst.agent.md b/.github/agents/analyst.agent.md
new file mode 100644
index 000000000..dfa4ca692
--- /dev/null
+++ b/.github/agents/analyst.agent.md
@@ -0,0 +1,15 @@
+# analyst.agent.md — Thin Wrapper
+#
+# Full agent definitions: .claude/agents/core/analyst.md
+
+## Role
+Analysis agent. Performs research, root-cause analysis, and pattern discovery.
+
+## Instructions
+- @.apm/instructions/project-rules.instructions.md
+- @.apm/prompts/research.prompt.md
+
+## Research Protocol (MANDATORY)
+1. claude-context search FIRST
+2. Verify with rg/ls
+3. Read docs/workflows/BASE_HANDBOOK.md
diff --git a/.github/agents/build-monitor.agent.md b/.github/agents/build-monitor.agent.md
new file mode 100644
index 000000000..2bb9c0910
--- /dev/null
+++ b/.github/agents/build-monitor.agent.md
@@ -0,0 +1,18 @@
+# build-monitor.agent.md — Thin Wrapper
+#
+# Full agent definitions: .claude/agents/build-monitor.md
+
+## Role
+Build monitoring agent. Validates Hugo builds, CI/CD pipelines, and performance.
+
+## Instructions
+- @.apm/instructions/project-rules.instructions.md
+
+## Gates
+- `bin/hugo-build` must pass (zero errors)
+- Chrome DevTools: zero console errors, zero 404s
+- `bin/rake test:critical` green on every commit
+
+## CI/CD
+- GitHub Actions → GitHub Pages
+- Workflows: .github/workflows/publish.yml, test.yml, sync-and-publish.yml
diff --git a/.github/agents/coder.agent.md b/.github/agents/coder.agent.md
new file mode 100644
index 000000000..7c8b2615e
--- /dev/null
+++ b/.github/agents/coder.agent.md
@@ -0,0 +1,26 @@
+# coder.agent.md — Thin Wrapper
+#
+# Canonical instructions: .apm/instructions/project-rules.instructions.md
+# Full agent definitions: .claude/agents/core/coder.md
+# 
+# APM compiles this agent for all target tools. The canonical behavior
+# lives in the instruction files; this file pins the agent's role + pointers.
+
+## Role
+Primary implementation agent. Writes code following TDD cycle (RED → GREEN → REFACTOR).
+
+## Instructions
+- @.apm/instructions/project-rules.instructions.md
+- @.apm/instructions/tdd-enforcement.instructions.md
+
+## For CSS/HTML work
+- @.apm/instructions/css-consolidation.instructions.md
+
+## For Ruby work
+- @docs/workflows/ruby.md
+
+## Constraints
+- Behavior-focused tests ONLY — reject structural/CSS class tests
+- Run `bin/rake test:critical` after every micro-change
+- Never use `npm` — use `bun`
+- Never create ad hoc `*.sh` test scripts
diff --git a/.github/agents/content-creator.agent.md b/.github/agents/content-creator.agent.md
new file mode 100644
index 000000000..e6c244ca8
--- /dev/null
+++ b/.github/agents/content-creator.agent.md
@@ -0,0 +1,20 @@
+# content-creator.agent.md — Thin Wrapper
+#
+# Canonical instructions: .apm/instructions/voice-guide.instructions.md
+# Full agent definitions: .claude/agents/content-creator.md
+
+## Role
+Content creation agent. Writes blog posts, LinkedIn posts, and marketing copy.
+
+## Instructions
+- @.apm/instructions/voice-guide.instructions.md
+- @.apm/instructions/icp-target.instructions.md
+- @docs/workflows/blog-pipeline.md
+- @docs/workflows/linkedin-post-pipeline.md
+
+## Constraints
+- Zero generic AI language — run slop-detector + humanizer pass
+- Cold-eyes gate (9 checks) mandatory before publishing
+- Every sentence needs a person doing something
+- Banned words: unlock, harness, leverage, dive in, game-changer, etc.
+- Use `-` not `—` for all dashes
diff --git a/.github/agents/content-marketer.agent.md b/.github/agents/content-marketer.agent.md
new file mode 100644
index 000000000..053ecd0a7
--- /dev/null
+++ b/.github/agents/content-marketer.agent.md
@@ -0,0 +1,10 @@
+# content-marketer.agent.md — Thin Wrapper
+#
+# Full agent definitions: .claude/agents/content/content-marketer.md
+
+## Role
+Content marketing agent. Plans content calendars, researches topics, audits SEO performance.
+
+## Instructions
+- @.apm/instructions/icp-target.instructions.md
+- @docs/projects/2510-seo-content-strategy/20-29-strategy/20.07-content-plan-icp-e-q2-2026.md
diff --git a/.github/agents/hugo-expert.agent.md b/.github/agents/hugo-expert.agent.md
new file mode 100644
index 000000000..24fe61525
--- /dev/null
+++ b/.github/agents/hugo-expert.agent.md
@@ -0,0 +1,25 @@
+# hugo-expert.agent.md — Thin Wrapper
+#
+# Full agent definitions: .claude/agents/hugo-expert.md
+
+## Role
+Hugo static site generator expert. Handles templates, shortcodes, content structure, and build optimization.
+
+## Instructions
+- @.apm/instructions/project-rules.instructions.md
+- @docs/workflows/css-consolidation.md
+
+## Key Files
+- `config/_default/hugo.toml` — Hugo configuration
+- `themes/beaver/layouts/` — Templates
+- `themes/beaver/assets/css/` — CSS (PostCSS pipeline)
+- `content/` — Content files (Markdown + Hugo frontmatter)
+
+## Commands
+- `bin/hugo-build` — Build + validate
+- `hugo server -D` — Dev server
+
+## Constraints
+- Never modify vendor CSS files
+- Preserve cascade order: Critical → Layout → Component → Theme → Footer
+- BEM methodology for all new CSS
diff --git a/.github/agents/planner.agent.md b/.github/agents/planner.agent.md
new file mode 100644
index 000000000..837cedb13
--- /dev/null
+++ b/.github/agents/planner.agent.md
@@ -0,0 +1,9 @@
+# planner.agent.md — Thin Wrapper
+#
+# Full agent definitions: .claude/agents/core/planner.md
+
+## Role
+Planning agent. Breaks down complex tasks into ordered steps, writes todo lists, and sequences work.
+
+## Instructions
+- @.apm/instructions/project-rules.instructions.md
diff --git a/.github/agents/qa-browser-tester.agent.md b/.github/agents/qa-browser-tester.agent.md
new file mode 100644
index 000000000..3a18ca639
--- /dev/null
+++ b/.github/agents/qa-browser-tester.agent.md
@@ -0,0 +1,16 @@
+# qa-browser-tester.agent.md — Thin Wrapper
+#
+# Full agent definitions: .claude/agents/validation/qa-browser-tester.md
+
+## Role
+QA browser tester. Runs Chrome DevTools validation after HTML/CSS/JS changes.
+
+## Instructions
+- @.apm/instructions/project-rules.instructions.md
+- @.apm/instructions/tdd-enforcement.instructions.md
+
+## Gates
+- Console: ZERO JS errors
+- Network: ZERO 404s for assets
+- Capture desktop + mobile screenshots
+- All gates pass → commit
diff --git a/.github/agents/researcher.agent.md b/.github/agents/researcher.agent.md
new file mode 100644
index 000000000..5d953465d
--- /dev/null
+++ b/.github/agents/researcher.agent.md
@@ -0,0 +1,10 @@
+# researcher.agent.md — Thin Wrapper
+#
+# Full agent definitions: .claude/agents/core/researcher.md
+
+## Role
+Research agent. Discovers existing patterns, framework docs, and best practices before implementation.
+
+## Instructions
+- @.apm/instructions/project-rules.instructions.md
+- @.apm/prompts/research.prompt.md
diff --git a/.github/agents/reviewer.agent.md b/.github/agents/reviewer.agent.md
new file mode 100644
index 000000000..989624346
--- /dev/null
+++ b/.github/agents/reviewer.agent.md
@@ -0,0 +1,17 @@
+# reviewer.agent.md — Thin Wrapper
+#
+# Canonical instructions: .apm/instructions/project-rules.instructions.md
+# Full agent definitions: .claude/agents/core/reviewer.md
+
+## Role
+Code review agent. Validates correctness, adherence to project rules, and TDD compliance.
+
+## Instructions
+- @.apm/instructions/project-rules.instructions.md
+- @.apm/instructions/tdd-enforcement.instructions.md
+
+## Gates
+- Zero structural test patterns (CSS class testing, HTML counting)
+- Visual regression tolerance verified
+- Four-eyes principle enforced for all critical changes
+- Feature branch + PR pattern for HTML/CSS/template changes
diff --git a/.github/agents/ruby-coder.agent.md b/.github/agents/ruby-coder.agent.md
new file mode 100644
index 000000000..510d186df
--- /dev/null
+++ b/.github/agents/ruby-coder.agent.md
@@ -0,0 +1,15 @@
+# ruby-coder.agent.md — Thin Wrapper
+#
+# Full agent definitions: .claude/agents/ruby/ruby-coder.md
+
+## Role
+Ruby implementation agent. Writes Ruby code (tests, sync scripts, lib/) following TDD.
+
+## Instructions
+- @.apm/instructions/project-rules.instructions.md
+- @.apm/instructions/tdd-enforcement.instructions.md
+- @docs/workflows/ruby.md
+
+## Commands
+- `bin/rake test:critical`
+- `bundle exec standardrb --fix`
diff --git a/.github/agents/ruby-reviewer.agent.md b/.github/agents/ruby-reviewer.agent.md
new file mode 100644
index 000000000..bdfcf27b8
--- /dev/null
+++ b/.github/agents/ruby-reviewer.agent.md
@@ -0,0 +1,14 @@
+# ruby-reviewer.agent.md — Thin Wrapper
+#
+# Full agent definitions: .claude/agents/ruby/ruby-reviewer.md
+
+## Role
+Ruby code reviewer. Validates Ruby code quality, TDD compliance, and Minitest patterns.
+
+## Instructions
+- @.apm/instructions/tdd-enforcement.instructions.md
+
+## Constraints
+- Behavior-focused tests ONLY — reject structural testing
+- StandardRB compliance
+- No ad hoc test scripts
diff --git a/.github/agents/screenshot-guardian.agent.md b/.github/agents/screenshot-guardian.agent.md
new file mode 100644
index 000000000..f160b3b75
--- /dev/null
+++ b/.github/agents/screenshot-guardian.agent.md
@@ -0,0 +1,15 @@
+# screenshot-guardian.agent.md — Thin Wrapper
+#
+# Full agent definitions: .claude/agents/core/screenshot-guardian.md
+
+## Role
+Visual regression guardian. Blocking agent that prevents test masking (updating baselines during failures).
+
+## Instructions
+- @.apm/instructions/tdd-enforcement.instructions.md
+- @.claude/hooks/pre-commit-screenshot-validation.md
+
+## Gates
+- ZERO tolerance: screenshot updates during visual test failures
+- Tolerance: 0.0 for refactors, ≤0.03 for new features
+- Both bin/test AND bin/dtest must pass before commit
diff --git a/.github/agents/seo-specialist.agent.md b/.github/agents/seo-specialist.agent.md
new file mode 100644
index 000000000..328a80718
--- /dev/null
+++ b/.github/agents/seo-specialist.agent.md
@@ -0,0 +1,21 @@
+# seo-specialist.agent.md — Thin Wrapper
+#
+# Full agent definitions: .claude/agents/seo-specialist.md
+
+## Role
+SEO optimization agent. Handles keyword research, meta tags, schema, and content optimization.
+
+## Instructions
+- @docs/90-99-content-strategy/seo-optimization/
+- @docs/90-99-content-strategy/strategy-analysis/
+
+## Key Files
+- `themes/beaver/layouts/partials/seo/enhanced-meta-tags.html` — SEO/og:image template
+- `themes/beaver/layouts/partials/seo/article-schema.html` — JSON-LD schema
+- `config/_default/hugo.toml` — Hugo config (permalinks, sitemap)
+
+## Constraints
+- Title ≤45 chars, description 150-160 chars
+- Primary keyword in first 100 words
+- 4+ internal links, 5+ external citations
+- No keyword stuffing
diff --git a/.github/agents/tester.agent.md b/.github/agents/tester.agent.md
new file mode 100644
index 000000000..8bc5f213b
--- /dev/null
+++ b/.github/agents/tester.agent.md
@@ -0,0 +1,25 @@
+# tester.agent.md — Thin Wrapper
+#
+# Canonical instructions: .apm/instructions/project-rules.instructions.md
+# Full agent definitions: .claude/agents/core/tester.md
+
+## Role
+Test agent. Writes and runs tests following TDD methodology.
+
+## Instructions
+- @.apm/instructions/project-rules.instructions.md
+- @.apm/instructions/tdd-enforcement.instructions.md
+
+## Framework
+- Minitest (test/system/, test/unit/)
+- capybara-screenshot-diff for visual regression
+
+## Commands
+- `bin/rake test:critical` — critical tests
+- `bin/test` — visual regression (macOS)
+- `bin/dtest` — visual regression (Linux/Docker)
+
+## Constraints
+- Behavior-focused tests ONLY
+- Assert shape, not config values
+- Run both bin/test AND bin/dtest before commit for HTML/CSS changes
diff --git a/.github/agents/tutorial-engineer.agent.md b/.github/agents/tutorial-engineer.agent.md
new file mode 100644
index 000000000..14374cb5e
--- /dev/null
+++ b/.github/agents/tutorial-engineer.agent.md
@@ -0,0 +1,10 @@
+# tutorial-engineer.agent.md — Thin Wrapper
+#
+# Full agent definitions: .claude/agents/content/tutorial-engineer.md
+
+## Role
+Tutorial engineer. Creates step-by-step how-to content, validates code examples, and ensures copy-paste-ability.
+
+## Instructions
+- @.apm/instructions/voice-guide.instructions.md
+- @.apm/instructions/icp-target.instructions.md
diff --git a/.github/agents/ui-problem-diagnosis.agent.md b/.github/agents/ui-problem-diagnosis.agent.md
new file mode 100644
index 000000000..9f88e69ee
--- /dev/null
+++ b/.github/agents/ui-problem-diagnosis.agent.md
@@ -0,0 +1,10 @@
+# ui-problem-diagnosis.agent.md — Thin Wrapper
+#
+# Full agent definitions: .claude/agents/validation/ui-problem-diagnosis-specialist.md
+
+## Role
+UI problem diagnosis specialist. Investigates visual regressions, cross-browser issues, and CSS bugs.
+
+## Instructions
+- @.apm/instructions/css-consolidation.instructions.md
+- @docs/projects/2509-css-migration/css-loading-order-analysis.md
diff --git a/.github/agents/ux-browser-validator.agent.md b/.github/agents/ux-browser-validator.agent.md
new file mode 100644
index 000000000..118ea4ffa
--- /dev/null
+++ b/.github/agents/ux-browser-validator.agent.md
@@ -0,0 +1,14 @@
+# ux-browser-validator.agent.md — Thin Wrapper
+#
+# Full agent definitions: .claude/agents/validation/ux-browser-validator.md
+
+## Role
+UX browser validator. Validates UX interactions, accessibility, and cross-browser behavior.
+
+## Instructions
+- @.apm/instructions/project-rules.instructions.md
+
+## Gates
+- WCAG 2.1 AA accessibility compliance
+- Semantic HTML elements (h1, button, main, nav)
+- Responsive design: desktop + mobile verification
diff --git a/.github/instructions/css-consolidation.instructions.md b/.github/instructions/css-consolidation.instructions.md
new file mode 100644
index 000000000..0f1faae2d
--- /dev/null
+++ b/.github/instructions/css-consolidation.instructions.md
@@ -0,0 +1,30 @@
+# CSS Consolidation Rules — JetThoughts Site
+#
+# MANDATORY for all CSS/HTML template changes.
+# Full docs: docs/workflows/css-consolidation.md
+# Project: docs/projects/2509-css-migration/css-loading-order-analysis.md
+
+## Architecture
+- **Methodology**: BEM (Block__Element--Modifier)
+- **Pipeline**: PostCSS (import, mixins, nested, autoprefixer, cssnano, purgecss)
+- **Cascade order**: Critical → Layout → Component → Theme → Footer
+
+## Rules
+- Never modify vendor files (`css/vendors/*.min.css`) or dynamic templates (`css/dynamic-*.css`)
+- Preserve cascade order in all changes
+- Extract whole rule sets; keep page-specific `.fl-node-{hash}` in original files
+- Run `bin/rake test:critical` after each extraction
+
+## FL-Node Extraction (Beaver Builder → BEM)
+- **Phase**: TDD RED-GREEN-REFACTOR cycle
+- **Validation**: `env PRECOMPILED_ASSETS=false bin/rake test:critical` after each change
+- **Pattern**: Add BEM classes first → verify visual fidelity → remove FL-nodes → retest
+- **Commit**: micro-commit on green tests only
+
+## Visual Regression
+- Tolerance: 0.0 for refactoring (zero visual changes)
+- Tolerance: ≤0.03 for new features only
+- Run BOTH `bin/test` and `bin/dtest` before commit
+- Visual failures are commit blockers, not warnings
+
+@see docs/workflows/css-consolidation.md
diff --git a/.github/instructions/icp-target.instructions.md b/.github/instructions/icp-target.instructions.md
new file mode 100644
index 000000000..e9457d522
--- /dev/null
+++ b/.github/instructions/icp-target.instructions.md
@@ -0,0 +1,46 @@
+# ICP Target — JetThoughts Primary Website Audience
+#
+# CANONICAL SOURCE: docs/90-99-content-strategy/strategy-analysis/90.10-icp-primary-website-target.md
+# This file is a COMPILED EXTRACT for APM — the full doc is authoritative.
+# Update the canonical source first, then sync this extract.
+#
+# MANDATORY reading before content, design, marketing, or blog work.
+# Compiled by APM into every AI tool's context.
+
+## ICP-E: Non-Technical Founder Burned by a Devshop
+
+### Who They Are
+- 35-50 years old
+- Non-technical or semi-technical
+- Raised funding ($500K–$2M seed)
+- Hired a dev shop or freelancers
+- Missed deadlines, buggy product, mounting technical debt
+- Starting to question: "Is it me? Is it them? How do I know?"
+
+### Their Pain
+- Can't evaluate code quality themselves
+- Dev shop says "almost done" for months
+- Every new feature breaks something else
+- No tests, no documentation, no deploy pipeline
+- Starting to lose investor confidence
+
+### What They Need
+- Someone who speaks plain English, not jargon
+- Proof that the problem is real and solvable (specific stories, not vague promises)
+- A clear path out: audit → stabilize → build right
+- Trust signals: named clients, specific results, honest trade-offs
+
+### What Resonates
+- Stories of other founders in the same situation
+- Specific numbers: "$40K wasted," "11-minute outage," "84 support tickets"
+- Honest acknowledgment that rescue is hard and expensive
+- The idea that good engineering is a competitive advantage
+- "You're not crazy — this IS what bad engineering looks like"
+
+### What Alienates
+- jargon-dense technical explanations before establishing trust
+- "We're the best" claims without evidence
+- Generic consulting language ("partner with you," "drive outcomes")
+- Ignoring the emotional reality (shame, frustration, fear of looking stupid)
+
+@see docs/90-99-content-strategy/strategy-analysis/90.10-icp-primary-website-target.md
diff --git a/.github/instructions/project-rules.instructions.md b/.github/instructions/project-rules.instructions.md
new file mode 100644
index 000000000..deb7313e4
--- /dev/null
+++ b/.github/instructions/project-rules.instructions.md
@@ -0,0 +1,79 @@
+# JetThoughts Site — Project Rules
+#
+# This is the SINGLE SOURCE for AI agent instructions across all tools.
+# APM compiles this to each tool's expected path (CLAUDE.md, .github/
+# copilot-instructions.md, etc.). Edit here, not in tool-specific files.
+#
+# CANONICAL SOURCE: THIS FILE. The existing CLAUDE.md is a Claude-specific
+# superset; over time CLI tool instructions converge here. Voice, ICP, TDD,
+# and CSS rules are split into separate instruction files for modularity.
+#
+# Full workflow docs: docs/workflows/
+# Full content docs: docs/90-99-content-strategy/
+
+## Stack
+- **Site**: Hugo 0.147.9 Extended static site generator
+- **Theme**: Custom "Beaver" theme (BEM + PostCSS)
+- **Tests**: Ruby 3.4, Minitest, Capybara, Selenium WebDriver
+- **CSS**: PostCSS pipeline (autoprefixer, cssnano, purgecss)
+- **Package managers**: Bun (JS, NOT npm), Bundler (Ruby)
+- **CI/CD**: GitHub Actions → GitHub Pages
+
+## Build & Test Commands
+- Build: `bin/hugo-build`
+- Test (critical): `bin/rake test:critical`
+- Test (visual/macOS): `bin/test`
+- Test (visual/Linux Docker): `bin/dtest`
+- Lint Ruby: `bundle exec standardrb --fix`
+
+## Research Protocol (MANDATORY — before any change)
+1. Search codebase with `claude-context` MCP FIRST
+2. Verify exact filenames with `rg`/`ls` as fallback
+3. Read `docs/workflows/BASE_HANDBOOK.md` at session start
+4. Read `docs/workflows/flow-router.md` to route to correct workflow
+5. For blog content: find existing posts via `docs/blog-post-index.md`
+6. Never guess slugs — verify with `ls content/blog/<slug>/index.md`
+
+## Non-Negotiables
+- **Zero generic AI language** in content — all AI-sounding phrases rejected
+- **Zero Hugo build breaks** — `bin/hugo-build` must pass for content changes
+- **Zero unsupported claims** — all assertions must have citations
+- **No ad hoc test scripts** — use `bin/rake test:critical` and `bin/test`
+- **No Python for analysis** — prefer `rg`, `sed`, `awk`, shell tools
+- **No duplicate throwaway files** — no `*_refactored.*`, `*_new.*`, `*_backup.*`
+- **New docs only under `docs/workflows/`** — else edit existing files
+- **Never commit coordinator/verification reports** — write to `/tmp/` instead
+
+## Content Work (Blog Posts, LinkedIn, Pages)
+- **BEFORE writing**: Read voice guide + ICP target (see instructions/voice-guide, instructions/icp-target)
+- **Blog posts**: Follow `docs/workflows/blog-pipeline.md` end-to-end
+- **LinkedIn**: Follow `docs/workflows/linkedin-post-pipeline.md`
+- **Cover images**: Follow `docs/workflows/cover-images.md` + `.stitch/design.md`
+- Run slop-detector + humanizer pass on all drafts
+- Cold-eyes gate (9 checks) mandatory before publishing
+
+## CSS / HTML Changes
+- **BEM methodology** mandatory for all new CSS
+- **FL-node extraction**: TDD cycle — RED → GREEN → REFACTOR → TEST
+- **Visual regression tolerance**: 0.0 for refactors, ≤0.03 for new features
+- **MANDATORY**: Run BOTH `bin/test` (macOS) AND `bin/dtest` (Linux Docker) before commit
+- **Feature branches**: HTML/CSS/template changes → feature branch → one PR per sprint
+- Zero tolerance: CSS class existence testing, framework selector usage in tests
+
+## TDD & Testing
+- **Cycle**: RED → GREEN → REFACTOR
+- **Framework**: Minitest (test/system/, test/unit/)
+- **Behavior-focused tests ONLY** — reject implementation/existence/config tests
+- **Avoid fragile config assertions** — assert shape, not tunable values
+- **Visual regression**: screenshot comparison, 0.0 tolerance for refactoring
+- Run `bin/rake test:critical` after every micro-change (<10 lines)
+- Chrome DevTools validation after HTML/CSS/JS changes: zero console errors, zero 404s
+
+## Behavioral Constraints
+- **User says "code is bad" or "over-engineered"**: HALT, 5-Why analysis, fix config, proceed
+- **Reflection triggers**: zero-defect tracking, test masking detection
+- **Four-eyes principle**: paired agents for all critical changes
+- **Direct-to-master only for**: content-only blog edits, commit fixes, docs, CLAUDE.md policy updates
+
+@see docs/workflows/ for full workflow documentation
+@see docs/90-99-content-strategy/ for content strategy
diff --git a/.github/instructions/tdd-enforcement.instructions.md b/.github/instructions/tdd-enforcement.instructions.md
new file mode 100644
index 000000000..a532cfa19
--- /dev/null
+++ b/.github/instructions/tdd-enforcement.instructions.md
@@ -0,0 +1,62 @@
+# TDD Enforcement — JetThoughts Site
+#
+# MANDATORY for all code changes. Extracted from CLAUDE.md TDD section.
+# Full docs: docs/60.03-tdd-quality-enforcement.md, docs/60.04-four-eyes-principle.md
+
+## TDD Cycle
+1. **RED**: Write a failing test that validates BEHAVIOR (not implementation)
+2. **GREEN**: Minimal code to make the test pass
+3. **REFACTOR**: Clean up while tests stay green
+4. **COMMIT**: Micro-commit on green tests
+
+## Test Quality Rules (ZERO TOLERANCE)
+
+### Prohibited Test Patterns — Agent MUST Reject
+- ❌ CSS class existence testing (`.fl-node-*`, `.has_css?(".px-4")`)
+- ❌ HTML structure counting (`page.all("div").count`)
+- ❌ Framework-generated selector usage (FL-Builder, Tailwind utility classes)
+- ❌ Method existence testing (`assert SomeClass.method_defined?`)
+- ❌ Config value assertions (hardcoded `q=90`, exact file sizes, specific dimensions)
+- ❌ Implementation detail testing
+
+### Required Test Patterns
+- ✅ Visual regression: `take_screenshot` → `assert_visual_match` with tolerance
+- ✅ User-visible content: `assert page.has_content?("text")`
+- ✅ Semantic elements: `assert page.has_css?("h1", text: "...")`
+- ✅ User interaction: `click_button` → `assert_current_path`
+- ✅ Shape assertions: `q=\d+`, `src contains wsrv.nl`, not `q=90`, not exact URLs
+
+### Avoid Fragile Config Assertions
+Don't hardcode tunable values. Assert the *shape*, not the configuration:
+- Bad: `assert_match /q=90/, src`
+- Good: `assert_match /q=\d+/, src`
+- Bad: `assert_match /w=360/, src`
+- Good: `assert has_css?("picture source")`
+
+## Visual Regression Gates (BLOCKING for HTML/CSS changes)
+- **Tolerance**: 0.0 for refactoring, ≤0.03 for new features
+- **Protocol**: Capture baseline BEFORE changes → compare AFTER → block on any diff >0%
+- **MANDATORY**: Run BOTH `bin/test` (macOS) AND `bin/dtest` (Linux Docker) before commit
+- Visual regression failures are COMMIT BLOCKERS, not warnings
+
+## Test Runner
+- Framework: Minitest
+- Critical tests: `bin/rake test:critical` (after every micro-change <10 lines)
+- Visual macOS: `bin/test`
+- Visual Linux/Docker: `bin/dtest`
+- NEVER create ad hoc `*.sh` test scripts
+
+## Chrome DevTools Validation (MANDATORY after HTML/CSS/JS)
+1. Start Hugo dev server, open page in Chrome DevTools
+2. Console: ZERO JS errors
+3. Network: ZERO 404s
+4. Capture desktop + mobile screenshots
+5. All gates pass → commit. Any gate fails → fix, repeat
+
+## Feature Branch + PR (HTML/CSS/template changes)
+1. `git checkout -b <sprint-name>`
+2. Multiple related commits on branch
+3. Both test gates green on each commit
+4. `git push -u origin <branch>`
+5. ONE PR per sprint via `gh pr create`
+6. Direct-to-master only for: content-only blog edits, commit fixes, docs, CLAUDE.md updates
diff --git a/.github/instructions/voice-guide.instructions.md b/.github/instructions/voice-guide.instructions.md
new file mode 100644
index 000000000..97aeec148
--- /dev/null
+++ b/.github/instructions/voice-guide.instructions.md
@@ -0,0 +1,68 @@
+# Voice Guide — JetThoughts Content
+#
+# CANONICAL SOURCE: docs/90-99-content-strategy/strategy-analysis/90.11-voice-guide.md
+# This file is a COMPILED EXTRACT for APM — the full doc is authoritative.
+# Update the canonical source first, then sync this extract.
+#
+# MANDATORY reading before writing ANY blog post, LinkedIn post, or marketing copy.
+# Compiled by APM into every AI tool's context.
+
+## Two Voices
+
+### Founder Posts (Paul Keen / JT leadership)
+- **Rob Walling directness** + **Rand Fishkin vulnerability**
+- Plain English, empathy, specific numbers
+- Talk to a non-technical founder burned by a devshop
+- "Been there" energy — practitioner not vendor
+- Trade-offs acknowledged honestly
+
+### Tech Posts (engineering team)
+- **Thoughtbot style**: open with tension, not features
+- Own your opinions — no "it depends" hedging
+- Code as narrative evidence, not decoration
+- Honest trade-offs: what does this solution fail to do?
+
+## Banned Words
+unlock, harness, leverage, dive in, game-changer, journey, seamless, robust,
+supercharge, revolutionize, embark, delve, discover, blast radius, suspect list
+
+## Banned Structural Patterns
+- Rule of three (parallel triads)
+- Signposting ("This post is about...", "Here's what you need to know")
+- Bold inline-header lists
+- Triple rhetorical questions
+- Slogany parallel closings ("X is A. Y is B.")
+- Therapist voice ("That's the hardest part")
+- Copula avoidance ("serves as" → use "is")
+- Impersonal fragment stacking — every sentence needs a human subject
+- Sustained staccato (3+ consecutive short fragments or uniform-length sentences)
+- Command structure repetition (3+ paragraphs starting with imperatives)
+- Apologetic caveats ("every project is different")
+- Telling instead of showing — replace adjectives with concrete mechanics
+- Fake authority generalizations ("We've seen this on every codebase")
+- Timeline fabrication — verify technology release dates before using "last year"
+- Use `-` not `—` for all dashes
+
+## Quality Tests (run on every paragraph)
+- **"Who" test**: who is doing what? If subject is a thing, rewrite with a person
+- **"Show" test**: did I show the mechanic or just label it?
+- **"Coffee" test**: would a tired founder talk like this at a bar?
+- **"90/10" test**: is ≥90% educational? Is promotion ≤10%?
+- **"Trade-off" test**: did I acknowledge what the solution fails to do?
+- **"Practitioner" test**: replace generalized scenarios with specific incidents
+
+## Mannerism Noun Cap
+discipline, cycle, pattern, canonical, workflow, framework, cadence, rhythm,
+contract, gate, gatekeeper, layer, round, review, pass
+→ Max 5 occurrences total per post. Vary to the concrete thing happening.
+
+## Cadence Quotas (per H2 section)
+- At least one sentence <9 words AND one >22 words
+- No 3 consecutive sentences within 3 words of each other
+- One single-sentence paragraph per H2
+- Max 3 sentences / 70 words per paragraph
+- No two sentences in same paragraph starting with same first word
+- "The..." paragraph openers: <20% of total paragraphs
+- "We..." paragraph openers: <15% of total paragraphs
+
+@see docs/90-99-content-strategy/strategy-analysis/90.11-voice-guide.md
diff --git a/.github/prompts/blog-pipeline.prompt.md b/.github/prompts/blog-pipeline.prompt.md
new file mode 100644
index 000000000..9ad0669b1
--- /dev/null
+++ b/.github/prompts/blog-pipeline.prompt.md
@@ -0,0 +1,75 @@
+# Blog Post Pipeline — Trigger Prompt
+#
+# Copy this prompt to write the next blog post from the content plan.
+# Full pipeline: docs/workflows/blog-pipeline.md
+#
+# This is a reusable PROMPT — APM compiles it to slash-commands for
+# each AI tool (e.g., /blog-post in Claude, #blog-post in Copilot).
+
+Write the next blog post following the full pipeline.
+
+## STEP 1 — Pick Topic
+Read the content plan: docs/projects/2510-seo-content-strategy/20-29-strategy/20.07-content-plan-icp-e-q2-2026.md
+Find the next 🔲 Planned topic. Confirm slug, primary keyword, target publish date.
+
+## STEP 2 — Read Context (mandatory, do not skip)
+- ICP: docs/90-99-content-strategy/strategy-analysis/90.10-icp-primary-website-target.md
+- Voice guide: docs/90-99-content-strategy/strategy-analysis/90.11-voice-guide.md
+- Content plan brief: read the topic's brief section
+- Design spec: .stitch/design.md
+
+## STEP 3 — Research
+- Web search for 5+ authoritative sources
+- Check existing posts for duplication: ls content/blog/ | grep -iE "<key-terms>"
+- Verify every statistic has a citable source
+
+## STEP 3b — Find Internal Posts to Reference
+- Use claude-context MCP search FIRST
+- For tag/slug lookups: docs/blog-post-index.md
+- Verify each slug: ls content/blog/<slug>/index.md
+- Collect at least 4 verified slugs before drafting
+
+## STEP 4 — Draft
+Create content/blog/<slug>/index.md with:
+- Frontmatter: title ≤45 chars, description 150-160 chars, date, slug, keywords, tags
+- Voice: plain English, empathy first, founder-to-founder tone
+- Structure: punchy hook → keyword in first 100 words → numbered H2s → "When NOT to" → JT proof → CTA → related reading → further reading
+- 4+ internal links in body, 5+ external citations
+- Paragraphs ≤3 sentences
+- Cut test: "what does the reader DO differently?" — delete if answer is "nothing new"
+
+## STEP 5 — Review Loop
+Spawn 3 parallel critic agents:
+1. Founder persona (Alex, burned CEO) — ≥8/10
+2. SEO/slop detector — ≥8/10 humanization
+3. Senior copy editor — "publish as-is" or "minor edits"
+
+## STEP 5c — Cold-Eyes Gate (mandatory)
+Spawn fresh-context reviewer with 9 explicit checks:
+1. Frontmatter vs body consistency
+2. Fluffy AI narration detection
+3. Fake authority stamps
+4. Timeline fabrication
+5. "Not X but Y" patterns
+6. Em dashes → hyphens
+7. Rule-of-three modifiers
+8. Cross-section template repetition
+9. Impersonal fragment stacking
+
+## STEP 6 — Cover Image
+- Read .stitch/design.md (6-slot layout)
+- Duplicate existing HTML from .stitch/designs/
+- Render: chrome-devtools → viewport 2400×1260×2 → screenshot
+- Downsample: magick → 2400×1260 → content/blog/<slug>/cover.png
+
+## STEP 7 — Validate
+- bin/hugo-build must pass (zero errors)
+- Chrome DevTools: zero console errors, zero 404s, og:image resolves
+
+## STEP 8 — Update Content Plan
+- Change topic status: 🔲 Planned → ✅ Published
+- Add commit SHA + changelog
+
+## STEP 9 — Commit
+- Stage: post directory + cover HTML + content plan update
+- Commit format: feat(content): add <topic> post
diff --git a/.github/prompts/cover-image.prompt.md b/.github/prompts/cover-image.prompt.md
new file mode 100644
index 000000000..2f53b5b4b
--- /dev/null
+++ b/.github/prompts/cover-image.prompt.md
@@ -0,0 +1,19 @@
+# Cover Image Generation — Trigger Prompt
+#
+# Reusable prompt for generating blog post cover images via Stitch design system.
+# Full spec: .stitch/design.md (canonical 6-slot layout)
+# Full workflow: docs/workflows/cover-images.md
+
+Generate a cover image for a blog post.
+
+## Steps
+1. Read .stitch/design.md for the 6-slot layout specification
+2. Duplicate an existing HTML from .stitch/designs/ (e.g., rails-8-1-active-job-continuations-cover.html)
+3. Change ONLY: title, category, pill, 3 headline lines, 3 chip labels+values, status text
+4. Render: chrome-devtools new_page → emulate viewport=2400x1260x2 → wait 2s → take_screenshot
+5. Downsample: magick {screenshot} -filter Lanczos -resize 2400x1260 -sampling-factor 4:4:4 -strip -quality 95 content/blog/<slug>/cover.png
+
+## Design System
+- JetVelocity: obsidian dark, Ruby red (#cc342d), neon purple (#a855f7)
+- Font: Inter (headlines), JetBrains Mono (code/metrics)
+- 6-slot grid layout — see .stitch/design.md for exact positioning
diff --git a/.github/prompts/research.prompt.md b/.github/prompts/research.prompt.md
new file mode 100644
index 000000000..76d9c1507
--- /dev/null
+++ b/.github/prompts/research.prompt.md
@@ -0,0 +1,26 @@
+# Research Protocol — Trigger Prompt
+#
+# Reusable prompt for research-first development.
+# MANDATORY sequence before any implementation.
+
+Research this topic before implementing: {TOPIC}
+
+## Research Sequence
+1. **Handbook search**: Search claude-context for existing patterns in /knowledge/ and docs/
+2. **Codebase search**: Search claude-context for similar implementations in this repo
+3. **Framework docs**: context7 for official documentation (Hugo, PostCSS, Ruby, etc.)
+4. **Package analysis**: package-search for dependency research (if using new packages)
+5. **Web search**: brave-search / searxng for current best practices (2025-2026)
+
+## Output Requirements
+- 5+ authoritative sources
+- Existing patterns found (or confirmation of none)
+- Framework documentation citations
+- Dependency evaluation (if applicable)
+- Recommendation with trade-offs
+
+## Anti-Patterns (BLOCKED)
+- Implementation before research
+- Single-source decisions
+- Dependencies installed without package-search verification
+- Slugs guessed without ls content/blog/ verification
diff --git a/.gitignore b/.gitignore
index 89d5f897c..8fc5b1277 100644
--- a/.gitignore
+++ b/.gitignore
@@ -32,6 +32,17 @@ hugo.linux
 !.stitch
 !.junie
 !.gitignore
+!.apm
+
+# APM-generated tool configs (committed — like package-lock.json)
+!.codex
+!.cursor
+!.gemini
+!.opencode
+!.windsurf
+!.kiro
+!.agents
+!.github
 
 # SnapDiff runtime
 *.diff.png
@@ -99,16 +110,28 @@ docs/projects/2509-css-migration/50-59-execution/
 .cache
 _workspace
 .claude/scheduled_tasks.lock
-.stitch/designs/
-.stitch/prototypes/
-.stitch/loop-status.md
-.stitch/next-prompt.md
-.stitch/stitch-mockup.html
-.stitch/SITE.md
-
-# Sprint coordinator/verification reports — working notes only, not project docs.
-# User policy 2026-04-30: write to /tmp/ instead, or restore-staged before commit.
-docs/projects/**/sprint-*-coordinator-report.md
-docs/projects/**/sprint-*-verification-report.md
-docs/projects/**/*-coordinator-report.md
-docs/projects/**/*-verification-report.md
+.stitch/designs/
+.stitch/prototypes/
+.stitch/loop-status.md
+.stitch/next-prompt.md
+.stitch/stitch-mockup.html
+.stitch/SITE.md
+
+# Sprint coordinator/verification reports — working notes only, not project docs.
+# User policy 2026-04-30: write to /tmp/ instead, or restore-staged before commit.
+docs/projects/**/sprint-*-coordinator-report.md
+docs/projects/**/sprint-*-verification-report.md
+docs/projects/**/*-coordinator-report.md
+docs/projects/**/*-verification-report.md
+.grepai/
+
+# APM (Agent Package Manager)
+# apm_modules/ is the local cache — like node_modules, do not commit
+# apm.lock.yaml IS committed — like package-lock.json, ensures reproducibility
+apm_modules/
+
+# Docker volumes (compose.mcp.yml)
+searxng_data/
+milvus_data/
+etcd_data/
+minio_data/
diff --git a/.junie/AGENTS.md b/.junie/AGENTS.md
index e860800d9..b422d1eb9 100644
--- a/.junie/AGENTS.md
+++ b/.junie/AGENTS.md
@@ -1,21 +1,13 @@
 # AGENTS.md — JetThoughts Blog (Junie)
+#
+# GENERATED BY APM — do not edit directly.
+# Source: .apm/instructions/project-rules.instructions.md (manual — Junie not yet an APM target)
+# Regenerate: apm install
 
-Canonical instructions live in `CLAUDE.md`. Read that first.
-Session start: always read `@docs/workflows/BASE_HANDBOOK.md` and `@docs/workflows/flow-router.md`.
+Canonical instructions: `.apm/instructions/project-rules.instructions.md`
+Canonical voice guide: `.apm/instructions/voice-guide.instructions.md`
+Canonical ICP target: `.apm/instructions/icp-target.instructions.md`
 
-**ICP (MANDATORY for content/design/marketing):** Read `docs/90-99-content-strategy/strategy-analysis/90.10-icp-primary-website-target.md` before any content or design work. Target: non-technical founder burned by a devshop.
+Session start: read `docs/workflows/BASE_HANDBOOK.md` and `docs/workflows/flow-router.md`.
 
-**Voice (MANDATORY when writing):** Read `docs/90-99-content-strategy/strategy-analysis/90.11-voice-guide.md`. Founder posts: Walling+Fishkin. Tech posts: thoughtbot style (tension-first, opinionated, code as evidence). Run stream-specific self-test.
-
-**Finding posts to reference:** Use claude-context MCP search: `Search the codebase at /Users/pftg/dev/jetthoughts.github.io for: "topic"`. Tag/slug index: `docs/blog-post-index.md`. Verify slugs: `ls content/blog/<slug>/index.md`.
-
-Workflow references:
-- `docs/workflows/css-consolidation.md`
-- `docs/workflows/blog-pipeline.md`
-- `docs/workflows/cover-images.md` (canonical spec in `.stitch/design.md`)
-- `docs/workflows/commands.md`
-- `docs/workflows/agents.md`
-
-Key commands:
-- `bin/hugo-build`
-- `bin/rake test:critical`
+Key commands: `bin/hugo-build` | `bin/rake test:critical`
diff --git a/.kiro/settings/mcp.json b/.kiro/settings/mcp.json
new file mode 100644
index 000000000..368800ab2
--- /dev/null
+++ b/.kiro/settings/mcp.json
@@ -0,0 +1,68 @@
+{
+  "mcpServers": {
+    "claude-context": {
+      "command": "npx",
+      "args": [
+        "-y",
+        "@anthropic/claude-context"
+      ]
+    },
+    "context7": {
+      "command": "npx",
+      "args": [
+        "-y",
+        "@upstash/context7-mcp"
+      ]
+    },
+    "brave-search": {
+      "command": "npx",
+      "env": {
+        "BRAVE_API_KEY": "${BRAVE_API_KEY}"
+      },
+      "args": [
+        "-y",
+        "@anthropic/brave-search-mcp"
+      ]
+    },
+    "lean-ctx": {
+      "command": "npx",
+      "args": [
+        "-y",
+        "lean-ctx-mcp"
+      ]
+    },
+    "chrome-devtools": {
+      "command": "npx",
+      "args": [
+        "-y",
+        "@anthropic/chrome-devtools-mcp"
+      ]
+    },
+    "package-search": {
+      "command": "npx",
+      "args": [
+        "-y",
+        "package-search-mcp"
+      ]
+    },
+    "github-mcp-server": {
+      "command": "npx",
+      "env": {
+        "GITHUB_TOKEN": "${GITHUB_TOKEN}"
+      },
+      "args": [
+        "-y",
+        "@anthropic/github-mcp-server"
+      ]
+    },
+    "searxng": {
+      "url": "http://localhost:8080"
+    },
+    "ollama": {
+      "url": "http://localhost:11434"
+    },
+    "milvus": {
+      "url": "http://localhost:19530"
+    }
+  }
+}
\ No newline at end of file
diff --git a/.kiro/skills/blog-post-pipeline/SKILL.md b/.kiro/skills/blog-post-pipeline/SKILL.md
new file mode 100644
index 000000000..5f2324f46
--- /dev/null
+++ b/.kiro/skills/blog-post-pipeline/SKILL.md
@@ -0,0 +1,41 @@
+# Blog Post Pipeline Skill
+#
+# Reusable 9-phase autonomous content pipeline for Hugo-based blogs.
+# This is a PRODUCER skill — published by jetthoughts/site for other repos.
+
+## Overview
+End-to-end pipeline from topic selection to published post with automated
+quality gates, multi-critic review, and cover image generation.
+
+## Phases
+1. Pick topic from content plan
+2. Read context (ICP, voice guide, brief)
+3. Research (5+ authoritative sources)
+4. Draft with voice enforcement + cadence quotas
+5. Review loop (3 parallel critic agents: founder, SEO/slop, editor)
+5c. Cold-eyes gate (9 explicit checks)
+6. Cover image (6-slot HTML → chrome-devtools render → lanczos downsample)
+7. Validate (Hugo build + Chrome DevTools)
+8. Update content plan
+9. Commit
+
+## Dependencies
+### MCP Servers
+- claude-context — semantic codebase search for existing posts
+- brave-search / searxng — web research
+- chrome-devtools — cover image rendering, validation
+
+### Other Skills
+- content-voice-enforcement — voice guide compliance
+- cover-image-generation — 6-slot layout rendering
+
+### Agents
+- content-creator — drafting
+- seo-specialist — SEO compliance
+- reviewer — quality gates
+
+## Trigger
+@.apm/prompts/blog-pipeline.prompt.md
+
+## Full Documentation
+docs/workflows/blog-pipeline.md
diff --git a/.kiro/skills/content-voice-enforcement/SKILL.md b/.kiro/skills/content-voice-enforcement/SKILL.md
new file mode 100644
index 000000000..df01df43f
--- /dev/null
+++ b/.kiro/skills/content-voice-enforcement/SKILL.md
@@ -0,0 +1,37 @@
+# Content Voice Enforcement Skill
+#
+# Reusable voice enforcement for JetThoughts content — blog posts, LinkedIn,
+# and marketing copy targeting the ICP-E audience (non-technical founder
+# burned by a devshop).
+
+## Overview
+Enforces the JetThoughts voice guide, banned word list, structural patterns,
+cadence quotas, and quality tests on all content.
+
+## Voice Guide
+See .apm/instructions/voice-guide.instructions.md for the full compiled extract.
+Canonical source: docs/90-99-content-strategy/strategy-analysis/90.11-voice-guide.md
+
+## Two Voices
+- **Founder posts**: Rob Walling directness + Rand Fishkin vulnerability
+- **Tech posts**: Thoughtbot style (tension-first, opinionated, code as evidence)
+
+## Quality Tests (run on every paragraph)
+- **"Who" test**: who is doing what?
+- **"Show" test**: did I show the mechanic or just label it?
+- **"Coffee" test**: would a tired founder talk like this at a bar?
+- **"90/10" test**: ≥90% educational, ≤10% promotional
+- **"Trade-off" test**: acknowledged what the solution fails to do?
+- **"Practitioner" test**: replace generalizations with specific incidents
+
+## Slop Detector + Cold-Eyes Gate
+- Run slop-detector on all drafts
+- Run humanizer pass if flagged
+- Cold-eyes gate (9 checks) mandatory before publishing
+- See docs/workflows/blog-pipeline.md STEP 5a + STEP 5c
+
+## Full Documentation
+.apm/instructions/voice-guide.instructions.md
+.apm/instructions/icp-target.instructions.md
+docs/90-99-content-strategy/strategy-analysis/90.11-voice-guide.md
+docs/90-99-content-strategy/strategy-analysis/90.10-icp-primary-website-target.md
diff --git a/.kiro/skills/cover-image-generation/SKILL.md b/.kiro/skills/cover-image-generation/SKILL.md
new file mode 100644
index 000000000..a367207dc
--- /dev/null
+++ b/.kiro/skills/cover-image-generation/SKILL.md
@@ -0,0 +1,32 @@
+# Cover Image Generation Skill
+#
+# Reusable skill for generating blog post cover images using the
+# JetVelocity 6-slot design system.
+
+## Overview
+Generates 2400×1260 cover images for blog posts using a 6-slot HTML layout
+rendered via chrome-devtools.
+
+## Design System
+- JetVelocity: obsidian dark, Ruby red (#cc342d), neon purple (#a855f7)
+- Font: Inter (headlines), JetBrains Mono (code/metrics)
+- 6-slot grid layout — canonical spec in .stitch/design.md
+
+## Workflow
+1. Read .stitch/design.md for the 6-slot layout specification
+2. Duplicate an existing HTML from .stitch/designs/
+3. Change ONLY: title, category, pill, 3 headline lines, 3 chip labels+values, status text
+4. Render: chrome-devtools new_page → emulate viewport=2400x1260x2 → wait 2s → take_screenshot
+5. Downsample: magick {screenshot} -filter Lanczos -resize 2400x1260 -sampling-factor 4:4:4 -strip -quality 95 content/blog/<slug>/cover.png
+6. Set frontmatter: cover_image: cover.png, cover_image_alt: descriptive alt text
+
+## Dependencies
+### MCP
+- chrome-devtools — rendering
+
+### System
+- ImageMagick (magick CLI) — downsampling
+
+## Full Documentation
+.stitch/design.md (canonical 6-slot spec)
+docs/workflows/cover-images.md
diff --git a/.kiro/skills/css-bem-migration/SKILL.md b/.kiro/skills/css-bem-migration/SKILL.md
new file mode 100644
index 000000000..9fcd1046f
--- /dev/null
+++ b/.kiro/skills/css-bem-migration/SKILL.md
@@ -0,0 +1,49 @@
+# CSS BEM Migration Skill
+#
+# Reusable methodology for migrating from FL-Builder (Beaver Builder)
+# auto-generated classes to BEM component architecture.
+
+## Overview
+TDD-driven FL-node extraction: add BEM classes → verify visual fidelity →
+remove FL-nodes → retest. Micro-commits on every green test.
+
+## Pattern
+1. RED: Identify FL-node CSS dependencies
+2. GREEN: Add BEM classes to HTML, consolidate CSS
+3. REFACTOR: Remove duplicate FL-node CSS, cleanup
+4. TEST: `env PRECOMPILED_ASSETS=false bin/rake test:critical`
+5. COMMIT: Micro-commit on green tests
+
+## BEM Component Structure
+```html
+<!-- Before (FL-node) -->
+<div class="fl-col fl-node-yx43bujcaiqn hover-box">
+
+<!-- Transition (BEM + FL retention) -->
+<div class="fl-col fl-col-small hover-box c-feature-card">
+
+<!-- Final (Pure BEM) -->
+<div class="c-feature-card c-feature-card--variant">
+```
+
+## CSS Consolidation Pattern
+```css
+/* Before: 3 separate FL-node selectors */
+.fl-node-yx43bujcaiqn { width: 32%; margin-right: 30px; }
+.fl-node-ktz4ipj39vd6 { width: 36%; margin: 0 30px; }
+.fl-node-m39uvorzy5g8 { width: 32%; margin-left: 30px; }
+
+/* After: DRY BEM pattern */
+.c-feature-card { width: 32%; margin-right: 30px; }
+.c-feature-card--second { width: 36%; margin: 0 30px; }
+.c-feature-card--third { margin-right: 0; }
+```
+
+## Constraints
+- Never modify vendor files or dynamic templates
+- Preserve cascade order: Critical → Layout → Component → Theme → Footer
+- Run BOTH bin/test (macOS) AND bin/dtest (Linux Docker) before commit
+
+## Full Documentation
+docs/workflows/css-consolidation.md
+docs/projects/2509-css-migration/css-loading-order-analysis.md
diff --git a/.kiro/skills/visual-regression-testing/SKILL.md b/.kiro/skills/visual-regression-testing/SKILL.md
new file mode 100644
index 000000000..ff12576f9
--- /dev/null
+++ b/.kiro/skills/visual-regression-testing/SKILL.md
@@ -0,0 +1,41 @@
+# Visual Regression Testing Skill
+#
+# Reusable visual regression testing protocol for Hugo static sites.
+
+## Overview
+Screenshot comparison-based visual testing using capybara-screenshot-diff.
+Blocks commits on any visual regression during refactoring.
+
+## Protocol
+1. Capture baseline screenshots BEFORE changes
+2. Make CSS/HTML changes
+3. Capture comparison screenshots AFTER changes
+4. Compare with tolerance:
+   - 0.0 for refactoring (zero visual changes)
+   - ≤0.03 for new features only
+5. Block commit on any difference > tolerance
+
+## Test Pattern
+```ruby
+def test_hero_section_visual_consistency
+  visit "/services/fractional-cto/"
+  screenshot_path = take_screenshot("fractional_cto_hero")
+  assert_visual_match(screenshot_path, "baseline/fractional_cto_hero.png",
+                      tolerance: 0.03)
+end
+```
+
+## Prohibited Patterns (ZERO TOLERANCE)
+- CSS class existence testing
+- HTML structure counting
+- Framework-generated selector usage
+- Method existence testing
+
+## Commands
+- `bin/test` — visual regression (macOS)
+- `bin/dtest` — visual regression (Linux/Docker, CI parity)
+- BOTH must pass before commit for HTML/CSS changes
+
+## Full Documentation
+.apm/instructions/tdd-enforcement.instructions.md
+docs/60.03-tdd-quality-enforcement.md
diff --git a/.kiro/steering/css-consolidation.md b/.kiro/steering/css-consolidation.md
new file mode 100644
index 000000000..15db1b509
--- /dev/null
+++ b/.kiro/steering/css-consolidation.md
@@ -0,0 +1,34 @@
+---
+inclusion: always
+---
+
+# CSS Consolidation Rules — JetThoughts Site
+#
+# MANDATORY for all CSS/HTML template changes.
+# Full docs: docs/workflows/css-consolidation.md
+# Project: docs/projects/2509-css-migration/css-loading-order-analysis.md
+
+## Architecture
+- **Methodology**: BEM (Block__Element--Modifier)
+- **Pipeline**: PostCSS (import, mixins, nested, autoprefixer, cssnano, purgecss)
+- **Cascade order**: Critical → Layout → Component → Theme → Footer
+
+## Rules
+- Never modify vendor files (`css/vendors/*.min.css`) or dynamic templates (`css/dynamic-*.css`)
+- Preserve cascade order in all changes
+- Extract whole rule sets; keep page-specific `.fl-node-{hash}` in original files
+- Run `bin/rake test:critical` after each extraction
+
+## FL-Node Extraction (Beaver Builder → BEM)
+- **Phase**: TDD RED-GREEN-REFACTOR cycle
+- **Validation**: `env PRECOMPILED_ASSETS=false bin/rake test:critical` after each change
+- **Pattern**: Add BEM classes first → verify visual fidelity → remove FL-nodes → retest
+- **Commit**: micro-commit on green tests only
+
+## Visual Regression
+- Tolerance: 0.0 for refactoring (zero visual changes)
+- Tolerance: ≤0.03 for new features only
+- Run BOTH `bin/test` and `bin/dtest` before commit
+- Visual failures are commit blockers, not warnings
+
+@see docs/workflows/css-consolidation.md
diff --git a/.kiro/steering/icp-target.md b/.kiro/steering/icp-target.md
new file mode 100644
index 000000000..edd88516b
--- /dev/null
+++ b/.kiro/steering/icp-target.md
@@ -0,0 +1,50 @@
+---
+inclusion: always
+---
+
+# ICP Target — JetThoughts Primary Website Audience
+#
+# CANONICAL SOURCE: docs/90-99-content-strategy/strategy-analysis/90.10-icp-primary-website-target.md
+# This file is a COMPILED EXTRACT for APM — the full doc is authoritative.
+# Update the canonical source first, then sync this extract.
+#
+# MANDATORY reading before content, design, marketing, or blog work.
+# Compiled by APM into every AI tool's context.
+
+## ICP-E: Non-Technical Founder Burned by a Devshop
+
+### Who They Are
+- 35-50 years old
+- Non-technical or semi-technical
+- Raised funding ($500K–$2M seed)
+- Hired a dev shop or freelancers
+- Missed deadlines, buggy product, mounting technical debt
+- Starting to question: "Is it me? Is it them? How do I know?"
+
+### Their Pain
+- Can't evaluate code quality themselves
+- Dev shop says "almost done" for months
+- Every new feature breaks something else
+- No tests, no documentation, no deploy pipeline
+- Starting to lose investor confidence
+
+### What They Need
+- Someone who speaks plain English, not jargon
+- Proof that the problem is real and solvable (specific stories, not vague promises)
+- A clear path out: audit → stabilize → build right
+- Trust signals: named clients, specific results, honest trade-offs
+
+### What Resonates
+- Stories of other founders in the same situation
+- Specific numbers: "$40K wasted," "11-minute outage," "84 support tickets"
+- Honest acknowledgment that rescue is hard and expensive
+- The idea that good engineering is a competitive advantage
+- "You're not crazy — this IS what bad engineering looks like"
+
+### What Alienates
+- jargon-dense technical explanations before establishing trust
+- "We're the best" claims without evidence
+- Generic consulting language ("partner with you," "drive outcomes")
+- Ignoring the emotional reality (shame, frustration, fear of looking stupid)
+
+@see docs/90-99-content-strategy/strategy-analysis/90.10-icp-primary-website-target.md
diff --git a/.kiro/steering/project-rules.md b/.kiro/steering/project-rules.md
new file mode 100644
index 000000000..3fed6b02a
--- /dev/null
+++ b/.kiro/steering/project-rules.md
@@ -0,0 +1,83 @@
+---
+inclusion: always
+---
+
+# JetThoughts Site — Project Rules
+#
+# This is the SINGLE SOURCE for AI agent instructions across all tools.
+# APM compiles this to each tool's expected path (CLAUDE.md, .github/
+# copilot-instructions.md, etc.). Edit here, not in tool-specific files.
+#
+# CANONICAL SOURCE: THIS FILE. The existing CLAUDE.md is a Claude-specific
+# superset; over time CLI tool instructions converge here. Voice, ICP, TDD,
+# and CSS rules are split into separate instruction files for modularity.
+#
+# Full workflow docs: docs/workflows/
+# Full content docs: docs/90-99-content-strategy/
+
+## Stack
+- **Site**: Hugo 0.147.9 Extended static site generator
+- **Theme**: Custom "Beaver" theme (BEM + PostCSS)
+- **Tests**: Ruby 3.4, Minitest, Capybara, Selenium WebDriver
+- **CSS**: PostCSS pipeline (autoprefixer, cssnano, purgecss)
+- **Package managers**: Bun (JS, NOT npm), Bundler (Ruby)
+- **CI/CD**: GitHub Actions → GitHub Pages
+
+## Build & Test Commands
+- Build: `bin/hugo-build`
+- Test (critical): `bin/rake test:critical`
+- Test (visual/macOS): `bin/test`
+- Test (visual/Linux Docker): `bin/dtest`
+- Lint Ruby: `bundle exec standardrb --fix`
+
+## Research Protocol (MANDATORY — before any change)
+1. Search codebase with `claude-context` MCP FIRST
+2. Verify exact filenames with `rg`/`ls` as fallback
+3. Read `docs/workflows/BASE_HANDBOOK.md` at session start
+4. Read `docs/workflows/flow-router.md` to route to correct workflow
+5. For blog content: find existing posts via `docs/blog-post-index.md`
+6. Never guess slugs — verify with `ls content/blog/<slug>/index.md`
+
+## Non-Negotiables
+- **Zero generic AI language** in content — all AI-sounding phrases rejected
+- **Zero Hugo build breaks** — `bin/hugo-build` must pass for content changes
+- **Zero unsupported claims** — all assertions must have citations
+- **No ad hoc test scripts** — use `bin/rake test:critical` and `bin/test`
+- **No Python for analysis** — prefer `rg`, `sed`, `awk`, shell tools
+- **No duplicate throwaway files** — no `*_refactored.*`, `*_new.*`, `*_backup.*`
+- **New docs only under `docs/workflows/`** — else edit existing files
+- **Never commit coordinator/verification reports** — write to `/tmp/` instead
+
+## Content Work (Blog Posts, LinkedIn, Pages)
+- **BEFORE writing**: Read voice guide + ICP target (see instructions/voice-guide, instructions/icp-target)
+- **Blog posts**: Follow `docs/workflows/blog-pipeline.md` end-to-end
+- **LinkedIn**: Follow `docs/workflows/linkedin-post-pipeline.md`
+- **Cover images**: Follow `docs/workflows/cover-images.md` + `.stitch/design.md`
+- Run slop-detector + humanizer pass on all drafts
+- Cold-eyes gate (9 checks) mandatory before publishing
+
+## CSS / HTML Changes
+- **BEM methodology** mandatory for all new CSS
+- **FL-node extraction**: TDD cycle — RED → GREEN → REFACTOR → TEST
+- **Visual regression tolerance**: 0.0 for refactors, ≤0.03 for new features
+- **MANDATORY**: Run BOTH `bin/test` (macOS) AND `bin/dtest` (Linux Docker) before commit
+- **Feature branches**: HTML/CSS/template changes → feature branch → one PR per sprint
+- Zero tolerance: CSS class existence testing, framework selector usage in tests
+
+## TDD & Testing
+- **Cycle**: RED → GREEN → REFACTOR
+- **Framework**: Minitest (test/system/, test/unit/)
+- **Behavior-focused tests ONLY** — reject implementation/existence/config tests
+- **Avoid fragile config assertions** — assert shape, not tunable values
+- **Visual regression**: screenshot comparison, 0.0 tolerance for refactoring
+- Run `bin/rake test:critical` after every micro-change (<10 lines)
+- Chrome DevTools validation after HTML/CSS/JS changes: zero console errors, zero 404s
+
+## Behavioral Constraints
+- **User says "code is bad" or "over-engineered"**: HALT, 5-Why analysis, fix config, proceed
+- **Reflection triggers**: zero-defect tracking, test masking detection
+- **Four-eyes principle**: paired agents for all critical changes
+- **Direct-to-master only for**: content-only blog edits, commit fixes, docs, CLAUDE.md policy updates
+
+@see docs/workflows/ for full workflow documentation
+@see docs/90-99-content-strategy/ for content strategy
diff --git a/.kiro/steering/tdd-enforcement.md b/.kiro/steering/tdd-enforcement.md
new file mode 100644
index 000000000..41f773d2d
--- /dev/null
+++ b/.kiro/steering/tdd-enforcement.md
@@ -0,0 +1,66 @@
+---
+inclusion: always
+---
+
+# TDD Enforcement — JetThoughts Site
+#
+# MANDATORY for all code changes. Extracted from CLAUDE.md TDD section.
+# Full docs: docs/60.03-tdd-quality-enforcement.md, docs/60.04-four-eyes-principle.md
+
+## TDD Cycle
+1. **RED**: Write a failing test that validates BEHAVIOR (not implementation)
+2. **GREEN**: Minimal code to make the test pass
+3. **REFACTOR**: Clean up while tests stay green
+4. **COMMIT**: Micro-commit on green tests
+
+## Test Quality Rules (ZERO TOLERANCE)
+
+### Prohibited Test Patterns — Agent MUST Reject
+- ❌ CSS class existence testing (`.fl-node-*`, `.has_css?(".px-4")`)
+- ❌ HTML structure counting (`page.all("div").count`)
+- ❌ Framework-generated selector usage (FL-Builder, Tailwind utility classes)
+- ❌ Method existence testing (`assert SomeClass.method_defined?`)
+- ❌ Config value assertions (hardcoded `q=90`, exact file sizes, specific dimensions)
+- ❌ Implementation detail testing
+
+### Required Test Patterns
+- ✅ Visual regression: `take_screenshot` → `assert_visual_match` with tolerance
+- ✅ User-visible content: `assert page.has_content?("text")`
+- ✅ Semantic elements: `assert page.has_css?("h1", text: "...")`
+- ✅ User interaction: `click_button` → `assert_current_path`
+- ✅ Shape assertions: `q=\d+`, `src contains wsrv.nl`, not `q=90`, not exact URLs
+
+### Avoid Fragile Config Assertions
+Don't hardcode tunable values. Assert the *shape*, not the configuration:
+- Bad: `assert_match /q=90/, src`
+- Good: `assert_match /q=\d+/, src`
+- Bad: `assert_match /w=360/, src`
+- Good: `assert has_css?("picture source")`
+
+## Visual Regression Gates (BLOCKING for HTML/CSS changes)
+- **Tolerance**: 0.0 for refactoring, ≤0.03 for new features
+- **Protocol**: Capture baseline BEFORE changes → compare AFTER → block on any diff >0%
+- **MANDATORY**: Run BOTH `bin/test` (macOS) AND `bin/dtest` (Linux Docker) before commit
+- Visual regression failures are COMMIT BLOCKERS, not warnings
+
+## Test Runner
+- Framework: Minitest
+- Critical tests: `bin/rake test:critical` (after every micro-change <10 lines)
+- Visual macOS: `bin/test`
+- Visual Linux/Docker: `bin/dtest`
+- NEVER create ad hoc `*.sh` test scripts
+
+## Chrome DevTools Validation (MANDATORY after HTML/CSS/JS)
+1. Start Hugo dev server, open page in Chrome DevTools
+2. Console: ZERO JS errors
+3. Network: ZERO 404s
+4. Capture desktop + mobile screenshots
+5. All gates pass → commit. Any gate fails → fix, repeat
+
+## Feature Branch + PR (HTML/CSS/template changes)
+1. `git checkout -b <sprint-name>`
+2. Multiple related commits on branch
+3. Both test gates green on each commit
+4. `git push -u origin <branch>`
+5. ONE PR per sprint via `gh pr create`
+6. Direct-to-master only for: content-only blog edits, commit fixes, docs, CLAUDE.md updates
diff --git a/.kiro/steering/voice-guide.md b/.kiro/steering/voice-guide.md
new file mode 100644
index 000000000..d610b703d
--- /dev/null
+++ b/.kiro/steering/voice-guide.md
@@ -0,0 +1,72 @@
+---
+inclusion: always
+---
+
+# Voice Guide — JetThoughts Content
+#
+# CANONICAL SOURCE: docs/90-99-content-strategy/strategy-analysis/90.11-voice-guide.md
+# This file is a COMPILED EXTRACT for APM — the full doc is authoritative.
+# Update the canonical source first, then sync this extract.
+#
+# MANDATORY reading before writing ANY blog post, LinkedIn post, or marketing copy.
+# Compiled by APM into every AI tool's context.
+
+## Two Voices
+
+### Founder Posts (Paul Keen / JT leadership)
+- **Rob Walling directness** + **Rand Fishkin vulnerability**
+- Plain English, empathy, specific numbers
+- Talk to a non-technical founder burned by a devshop
+- "Been there" energy — practitioner not vendor
+- Trade-offs acknowledged honestly
+
+### Tech Posts (engineering team)
+- **Thoughtbot style**: open with tension, not features
+- Own your opinions — no "it depends" hedging
+- Code as narrative evidence, not decoration
+- Honest trade-offs: what does this solution fail to do?
+
+## Banned Words
+unlock, harness, leverage, dive in, game-changer, journey, seamless, robust,
+supercharge, revolutionize, embark, delve, discover, blast radius, suspect list
+
+## Banned Structural Patterns
+- Rule of three (parallel triads)
+- Signposting ("This post is about...", "Here's what you need to know")
+- Bold inline-header lists
+- Triple rhetorical questions
+- Slogany parallel closings ("X is A. Y is B.")
+- Therapist voice ("That's the hardest part")
+- Copula avoidance ("serves as" → use "is")
+- Impersonal fragment stacking — every sentence needs a human subject
+- Sustained staccato (3+ consecutive short fragments or uniform-length sentences)
+- Command structure repetition (3+ paragraphs starting with imperatives)
+- Apologetic caveats ("every project is different")
+- Telling instead of showing — replace adjectives with concrete mechanics
+- Fake authority generalizations ("We've seen this on every codebase")
+- Timeline fabrication — verify technology release dates before using "last year"
+- Use `-` not `—` for all dashes
+
+## Quality Tests (run on every paragraph)
+- **"Who" test**: who is doing what? If subject is a thing, rewrite with a person
+- **"Show" test**: did I show the mechanic or just label it?
+- **"Coffee" test**: would a tired founder talk like this at a bar?
+- **"90/10" test**: is ≥90% educational? Is promotion ≤10%?
+- **"Trade-off" test**: did I acknowledge what the solution fails to do?
+- **"Practitioner" test**: replace generalized scenarios with specific incidents
+
+## Mannerism Noun Cap
+discipline, cycle, pattern, canonical, workflow, framework, cadence, rhythm,
+contract, gate, gatekeeper, layer, round, review, pass
+→ Max 5 occurrences total per post. Vary to the concrete thing happening.
+
+## Cadence Quotas (per H2 section)
+- At least one sentence <9 words AND one >22 words
+- No 3 consecutive sentences within 3 words of each other
+- One single-sentence paragraph per H2
+- Max 3 sentences / 70 words per paragraph
+- No two sentences in same paragraph starting with same first word
+- "The..." paragraph openers: <20% of total paragraphs
+- "We..." paragraph openers: <15% of total paragraphs
+
+@see docs/90-99-content-strategy/strategy-analysis/90.11-voice-guide.md
diff --git a/.opencode/agents/analyst.md b/.opencode/agents/analyst.md
new file mode 100644
index 000000000..dfa4ca692
--- /dev/null
+++ b/.opencode/agents/analyst.md
@@ -0,0 +1,15 @@
+# analyst.agent.md — Thin Wrapper
+#
+# Full agent definitions: .claude/agents/core/analyst.md
+
+## Role
+Analysis agent. Performs research, root-cause analysis, and pattern discovery.
+
+## Instructions
+- @.apm/instructions/project-rules.instructions.md
+- @.apm/prompts/research.prompt.md
+
+## Research Protocol (MANDATORY)
+1. claude-context search FIRST
+2. Verify with rg/ls
+3. Read docs/workflows/BASE_HANDBOOK.md
diff --git a/.opencode/agents/build-monitor.md b/.opencode/agents/build-monitor.md
new file mode 100644
index 000000000..2bb9c0910
--- /dev/null
+++ b/.opencode/agents/build-monitor.md
@@ -0,0 +1,18 @@
+# build-monitor.agent.md — Thin Wrapper
+#
+# Full agent definitions: .claude/agents/build-monitor.md
+
+## Role
+Build monitoring agent. Validates Hugo builds, CI/CD pipelines, and performance.
+
+## Instructions
+- @.apm/instructions/project-rules.instructions.md
+
+## Gates
+- `bin/hugo-build` must pass (zero errors)
+- Chrome DevTools: zero console errors, zero 404s
+- `bin/rake test:critical` green on every commit
+
+## CI/CD
+- GitHub Actions → GitHub Pages
+- Workflows: .github/workflows/publish.yml, test.yml, sync-and-publish.yml
diff --git a/.opencode/agents/coder.md b/.opencode/agents/coder.md
new file mode 100644
index 000000000..7c8b2615e
--- /dev/null
+++ b/.opencode/agents/coder.md
@@ -0,0 +1,26 @@
+# coder.agent.md — Thin Wrapper
+#
+# Canonical instructions: .apm/instructions/project-rules.instructions.md
+# Full agent definitions: .claude/agents/core/coder.md
+# 
+# APM compiles this agent for all target tools. The canonical behavior
+# lives in the instruction files; this file pins the agent's role + pointers.
+
+## Role
+Primary implementation agent. Writes code following TDD cycle (RED → GREEN → REFACTOR).
+
+## Instructions
+- @.apm/instructions/project-rules.instructions.md
+- @.apm/instructions/tdd-enforcement.instructions.md
+
+## For CSS/HTML work
+- @.apm/instructions/css-consolidation.instructions.md
+
+## For Ruby work
+- @docs/workflows/ruby.md
+
+## Constraints
+- Behavior-focused tests ONLY — reject structural/CSS class tests
+- Run `bin/rake test:critical` after every micro-change
+- Never use `npm` — use `bun`
+- Never create ad hoc `*.sh` test scripts
diff --git a/.opencode/agents/content-creator.md b/.opencode/agents/content-creator.md
new file mode 100644
index 000000000..e6c244ca8
--- /dev/null
+++ b/.opencode/agents/content-creator.md
@@ -0,0 +1,20 @@
+# content-creator.agent.md — Thin Wrapper
+#
+# Canonical instructions: .apm/instructions/voice-guide.instructions.md
+# Full agent definitions: .claude/agents/content-creator.md
+
+## Role
+Content creation agent. Writes blog posts, LinkedIn posts, and marketing copy.
+
+## Instructions
+- @.apm/instructions/voice-guide.instructions.md
+- @.apm/instructions/icp-target.instructions.md
+- @docs/workflows/blog-pipeline.md
+- @docs/workflows/linkedin-post-pipeline.md
+
+## Constraints
+- Zero generic AI language — run slop-detector + humanizer pass
+- Cold-eyes gate (9 checks) mandatory before publishing
+- Every sentence needs a person doing something
+- Banned words: unlock, harness, leverage, dive in, game-changer, etc.
+- Use `-` not `—` for all dashes
diff --git a/.opencode/agents/content-marketer.md b/.opencode/agents/content-marketer.md
new file mode 100644
index 000000000..053ecd0a7
--- /dev/null
+++ b/.opencode/agents/content-marketer.md
@@ -0,0 +1,10 @@
+# content-marketer.agent.md — Thin Wrapper
+#
+# Full agent definitions: .claude/agents/content/content-marketer.md
+
+## Role
+Content marketing agent. Plans content calendars, researches topics, audits SEO performance.
+
+## Instructions
+- @.apm/instructions/icp-target.instructions.md
+- @docs/projects/2510-seo-content-strategy/20-29-strategy/20.07-content-plan-icp-e-q2-2026.md
diff --git a/.opencode/agents/hugo-expert.md b/.opencode/agents/hugo-expert.md
new file mode 100644
index 000000000..24fe61525
--- /dev/null
+++ b/.opencode/agents/hugo-expert.md
@@ -0,0 +1,25 @@
+# hugo-expert.agent.md — Thin Wrapper
+#
+# Full agent definitions: .claude/agents/hugo-expert.md
+
+## Role
+Hugo static site generator expert. Handles templates, shortcodes, content structure, and build optimization.
+
+## Instructions
+- @.apm/instructions/project-rules.instructions.md
+- @docs/workflows/css-consolidation.md
+
+## Key Files
+- `config/_default/hugo.toml` — Hugo configuration
+- `themes/beaver/layouts/` — Templates
+- `themes/beaver/assets/css/` — CSS (PostCSS pipeline)
+- `content/` — Content files (Markdown + Hugo frontmatter)
+
+## Commands
+- `bin/hugo-build` — Build + validate
+- `hugo server -D` — Dev server
+
+## Constraints
+- Never modify vendor CSS files
+- Preserve cascade order: Critical → Layout → Component → Theme → Footer
+- BEM methodology for all new CSS
diff --git a/.opencode/agents/planner.md b/.opencode/agents/planner.md
new file mode 100644
index 000000000..837cedb13
--- /dev/null
+++ b/.opencode/agents/planner.md
@@ -0,0 +1,9 @@
+# planner.agent.md — Thin Wrapper
+#
+# Full agent definitions: .claude/agents/core/planner.md
+
+## Role
+Planning agent. Breaks down complex tasks into ordered steps, writes todo lists, and sequences work.
+
+## Instructions
+- @.apm/instructions/project-rules.instructions.md
diff --git a/.opencode/agents/qa-browser-tester.md b/.opencode/agents/qa-browser-tester.md
new file mode 100644
index 000000000..3a18ca639
--- /dev/null
+++ b/.opencode/agents/qa-browser-tester.md
@@ -0,0 +1,16 @@
+# qa-browser-tester.agent.md — Thin Wrapper
+#
+# Full agent definitions: .claude/agents/validation/qa-browser-tester.md
+
+## Role
+QA browser tester. Runs Chrome DevTools validation after HTML/CSS/JS changes.
+
+## Instructions
+- @.apm/instructions/project-rules.instructions.md
+- @.apm/instructions/tdd-enforcement.instructions.md
+
+## Gates
+- Console: ZERO JS errors
+- Network: ZERO 404s for assets
+- Capture desktop + mobile screenshots
+- All gates pass → commit
diff --git a/.opencode/agents/researcher.md b/.opencode/agents/researcher.md
new file mode 100644
index 000000000..5d953465d
--- /dev/null
+++ b/.opencode/agents/researcher.md
@@ -0,0 +1,10 @@
+# researcher.agent.md — Thin Wrapper
+#
+# Full agent definitions: .claude/agents/core/researcher.md
+
+## Role
+Research agent. Discovers existing patterns, framework docs, and best practices before implementation.
+
+## Instructions
+- @.apm/instructions/project-rules.instructions.md
+- @.apm/prompts/research.prompt.md
diff --git a/.opencode/agents/reviewer.md b/.opencode/agents/reviewer.md
new file mode 100644
index 000000000..989624346
--- /dev/null
+++ b/.opencode/agents/reviewer.md
@@ -0,0 +1,17 @@
+# reviewer.agent.md — Thin Wrapper
+#
+# Canonical instructions: .apm/instructions/project-rules.instructions.md
+# Full agent definitions: .claude/agents/core/reviewer.md
+
+## Role
+Code review agent. Validates correctness, adherence to project rules, and TDD compliance.
+
+## Instructions
+- @.apm/instructions/project-rules.instructions.md
+- @.apm/instructions/tdd-enforcement.instructions.md
+
+## Gates
+- Zero structural test patterns (CSS class testing, HTML counting)
+- Visual regression tolerance verified
+- Four-eyes principle enforced for all critical changes
+- Feature branch + PR pattern for HTML/CSS/template changes
diff --git a/.opencode/agents/ruby-coder.md b/.opencode/agents/ruby-coder.md
new file mode 100644
index 000000000..510d186df
--- /dev/null
+++ b/.opencode/agents/ruby-coder.md
@@ -0,0 +1,15 @@
+# ruby-coder.agent.md — Thin Wrapper
+#
+# Full agent definitions: .claude/agents/ruby/ruby-coder.md
+
+## Role
+Ruby implementation agent. Writes Ruby code (tests, sync scripts, lib/) following TDD.
+
+## Instructions
+- @.apm/instructions/project-rules.instructions.md
+- @.apm/instructions/tdd-enforcement.instructions.md
+- @docs/workflows/ruby.md
+
+## Commands
+- `bin/rake test:critical`
+- `bundle exec standardrb --fix`
diff --git a/.opencode/agents/ruby-reviewer.md b/.opencode/agents/ruby-reviewer.md
new file mode 100644
index 000000000..bdfcf27b8
--- /dev/null
+++ b/.opencode/agents/ruby-reviewer.md
@@ -0,0 +1,14 @@
+# ruby-reviewer.agent.md — Thin Wrapper
+#
+# Full agent definitions: .claude/agents/ruby/ruby-reviewer.md
+
+## Role
+Ruby code reviewer. Validates Ruby code quality, TDD compliance, and Minitest patterns.
+
+## Instructions
+- @.apm/instructions/tdd-enforcement.instructions.md
+
+## Constraints
+- Behavior-focused tests ONLY — reject structural testing
+- StandardRB compliance
+- No ad hoc test scripts
diff --git a/.opencode/agents/screenshot-guardian.md b/.opencode/agents/screenshot-guardian.md
new file mode 100644
index 000000000..f160b3b75
--- /dev/null
+++ b/.opencode/agents/screenshot-guardian.md
@@ -0,0 +1,15 @@
+# screenshot-guardian.agent.md — Thin Wrapper
+#
+# Full agent definitions: .claude/agents/core/screenshot-guardian.md
+
+## Role
+Visual regression guardian. Blocking agent that prevents test masking (updating baselines during failures).
+
+## Instructions
+- @.apm/instructions/tdd-enforcement.instructions.md
+- @.claude/hooks/pre-commit-screenshot-validation.md
+
+## Gates
+- ZERO tolerance: screenshot updates during visual test failures
+- Tolerance: 0.0 for refactors, ≤0.03 for new features
+- Both bin/test AND bin/dtest must pass before commit
diff --git a/.opencode/agents/seo-specialist.md b/.opencode/agents/seo-specialist.md
new file mode 100644
index 000000000..328a80718
--- /dev/null
+++ b/.opencode/agents/seo-specialist.md
@@ -0,0 +1,21 @@
+# seo-specialist.agent.md — Thin Wrapper
+#
+# Full agent definitions: .claude/agents/seo-specialist.md
+
+## Role
+SEO optimization agent. Handles keyword research, meta tags, schema, and content optimization.
+
+## Instructions
+- @docs/90-99-content-strategy/seo-optimization/
+- @docs/90-99-content-strategy/strategy-analysis/
+
+## Key Files
+- `themes/beaver/layouts/partials/seo/enhanced-meta-tags.html` — SEO/og:image template
+- `themes/beaver/layouts/partials/seo/article-schema.html` — JSON-LD schema
+- `config/_default/hugo.toml` — Hugo config (permalinks, sitemap)
+
+## Constraints
+- Title ≤45 chars, description 150-160 chars
+- Primary keyword in first 100 words
+- 4+ internal links, 5+ external citations
+- No keyword stuffing
diff --git a/.opencode/agents/tester.md b/.opencode/agents/tester.md
new file mode 100644
index 000000000..8bc5f213b
--- /dev/null
+++ b/.opencode/agents/tester.md
@@ -0,0 +1,25 @@
+# tester.agent.md — Thin Wrapper
+#
+# Canonical instructions: .apm/instructions/project-rules.instructions.md
+# Full agent definitions: .claude/agents/core/tester.md
+
+## Role
+Test agent. Writes and runs tests following TDD methodology.
+
+## Instructions
+- @.apm/instructions/project-rules.instructions.md
+- @.apm/instructions/tdd-enforcement.instructions.md
+
+## Framework
+- Minitest (test/system/, test/unit/)
+- capybara-screenshot-diff for visual regression
+
+## Commands
+- `bin/rake test:critical` — critical tests
+- `bin/test` — visual regression (macOS)
+- `bin/dtest` — visual regression (Linux/Docker)
+
+## Constraints
+- Behavior-focused tests ONLY
+- Assert shape, not config values
+- Run both bin/test AND bin/dtest before commit for HTML/CSS changes
diff --git a/.opencode/agents/tutorial-engineer.md b/.opencode/agents/tutorial-engineer.md
new file mode 100644
index 000000000..14374cb5e
--- /dev/null
+++ b/.opencode/agents/tutorial-engineer.md
@@ -0,0 +1,10 @@
+# tutorial-engineer.agent.md — Thin Wrapper
+#
+# Full agent definitions: .claude/agents/content/tutorial-engineer.md
+
+## Role
+Tutorial engineer. Creates step-by-step how-to content, validates code examples, and ensures copy-paste-ability.
+
+## Instructions
+- @.apm/instructions/voice-guide.instructions.md
+- @.apm/instructions/icp-target.instructions.md
diff --git a/.opencode/agents/ui-problem-diagnosis.md b/.opencode/agents/ui-problem-diagnosis.md
new file mode 100644
index 000000000..9f88e69ee
--- /dev/null
+++ b/.opencode/agents/ui-problem-diagnosis.md
@@ -0,0 +1,10 @@
+# ui-problem-diagnosis.agent.md — Thin Wrapper
+#
+# Full agent definitions: .claude/agents/validation/ui-problem-diagnosis-specialist.md
+
+## Role
+UI problem diagnosis specialist. Investigates visual regressions, cross-browser issues, and CSS bugs.
+
+## Instructions
+- @.apm/instructions/css-consolidation.instructions.md
+- @docs/projects/2509-css-migration/css-loading-order-analysis.md
diff --git a/.opencode/agents/ux-browser-validator.md b/.opencode/agents/ux-browser-validator.md
new file mode 100644
index 000000000..118ea4ffa
--- /dev/null
+++ b/.opencode/agents/ux-browser-validator.md
@@ -0,0 +1,14 @@
+# ux-browser-validator.agent.md — Thin Wrapper
+#
+# Full agent definitions: .claude/agents/validation/ux-browser-validator.md
+
+## Role
+UX browser validator. Validates UX interactions, accessibility, and cross-browser behavior.
+
+## Instructions
+- @.apm/instructions/project-rules.instructions.md
+
+## Gates
+- WCAG 2.1 AA accessibility compliance
+- Semantic HTML elements (h1, button, main, nav)
+- Responsive design: desktop + mobile verification
diff --git a/.opencode/commands/blog-pipeline.md b/.opencode/commands/blog-pipeline.md
new file mode 100644
index 000000000..5c9dae536
--- /dev/null
+++ b/.opencode/commands/blog-pipeline.md
@@ -0,0 +1,79 @@
+---
+{}
+---
+
+# Blog Post Pipeline — Trigger Prompt
+#
+# Copy this prompt to write the next blog post from the content plan.
+# Full pipeline: docs/workflows/blog-pipeline.md
+#
+# This is a reusable PROMPT — APM compiles it to slash-commands for
+# each AI tool (e.g., /blog-post in Claude, #blog-post in Copilot).
+
+Write the next blog post following the full pipeline.
+
+## STEP 1 — Pick Topic
+Read the content plan: docs/projects/2510-seo-content-strategy/20-29-strategy/20.07-content-plan-icp-e-q2-2026.md
+Find the next 🔲 Planned topic. Confirm slug, primary keyword, target publish date.
+
+## STEP 2 — Read Context (mandatory, do not skip)
+- ICP: docs/90-99-content-strategy/strategy-analysis/90.10-icp-primary-website-target.md
+- Voice guide: docs/90-99-content-strategy/strategy-analysis/90.11-voice-guide.md
+- Content plan brief: read the topic's brief section
+- Design spec: .stitch/design.md
+
+## STEP 3 — Research
+- Web search for 5+ authoritative sources
+- Check existing posts for duplication: ls content/blog/ | grep -iE "<key-terms>"
+- Verify every statistic has a citable source
+
+## STEP 3b — Find Internal Posts to Reference
+- Use claude-context MCP search FIRST
+- For tag/slug lookups: docs/blog-post-index.md
+- Verify each slug: ls content/blog/<slug>/index.md
+- Collect at least 4 verified slugs before drafting
+
+## STEP 4 — Draft
+Create content/blog/<slug>/index.md with:
+- Frontmatter: title ≤45 chars, description 150-160 chars, date, slug, keywords, tags
+- Voice: plain English, empathy first, founder-to-founder tone
+- Structure: punchy hook → keyword in first 100 words → numbered H2s → "When NOT to" → JT proof → CTA → related reading → further reading
+- 4+ internal links in body, 5+ external citations
+- Paragraphs ≤3 sentences
+- Cut test: "what does the reader DO differently?" — delete if answer is "nothing new"
+
+## STEP 5 — Review Loop
+Spawn 3 parallel critic agents:
+1. Founder persona (Alex, burned CEO) — ≥8/10
+2. SEO/slop detector — ≥8/10 humanization
+3. Senior copy editor — "publish as-is" or "minor edits"
+
+## STEP 5c — Cold-Eyes Gate (mandatory)
+Spawn fresh-context reviewer with 9 explicit checks:
+1. Frontmatter vs body consistency
+2. Fluffy AI narration detection
+3. Fake authority stamps
+4. Timeline fabrication
+5. "Not X but Y" patterns
+6. Em dashes → hyphens
+7. Rule-of-three modifiers
+8. Cross-section template repetition
+9. Impersonal fragment stacking
+
+## STEP 6 — Cover Image
+- Read .stitch/design.md (6-slot layout)
+- Duplicate existing HTML from .stitch/designs/
+- Render: chrome-devtools → viewport 2400×1260×2 → screenshot
+- Downsample: magick → 2400×1260 → content/blog/<slug>/cover.png
+
+## STEP 7 — Validate
+- bin/hugo-build must pass (zero errors)
+- Chrome DevTools: zero console errors, zero 404s, og:image resolves
+
+## STEP 8 — Update Content Plan
+- Change topic status: 🔲 Planned → ✅ Published
+- Add commit SHA + changelog
+
+## STEP 9 — Commit
+- Stage: post directory + cover HTML + content plan update
+- Commit format: feat(content): add <topic> post
\ No newline at end of file
diff --git a/.opencode/commands/cover-image.md b/.opencode/commands/cover-image.md
new file mode 100644
index 000000000..024cd8573
--- /dev/null
+++ b/.opencode/commands/cover-image.md
@@ -0,0 +1,23 @@
+---
+{}
+---
+
+# Cover Image Generation — Trigger Prompt
+#
+# Reusable prompt for generating blog post cover images via Stitch design system.
+# Full spec: .stitch/design.md (canonical 6-slot layout)
+# Full workflow: docs/workflows/cover-images.md
+
+Generate a cover image for a blog post.
+
+## Steps
+1. Read .stitch/design.md for the 6-slot layout specification
+2. Duplicate an existing HTML from .stitch/designs/ (e.g., rails-8-1-active-job-continuations-cover.html)
+3. Change ONLY: title, category, pill, 3 headline lines, 3 chip labels+values, status text
+4. Render: chrome-devtools new_page → emulate viewport=2400x1260x2 → wait 2s → take_screenshot
+5. Downsample: magick {screenshot} -filter Lanczos -resize 2400x1260 -sampling-factor 4:4:4 -strip -quality 95 content/blog/<slug>/cover.png
+
+## Design System
+- JetVelocity: obsidian dark, Ruby red (#cc342d), neon purple (#a855f7)
+- Font: Inter (headlines), JetBrains Mono (code/metrics)
+- 6-slot grid layout — see .stitch/design.md for exact positioning
\ No newline at end of file
diff --git a/.opencode/commands/research.md b/.opencode/commands/research.md
new file mode 100644
index 000000000..59101394b
--- /dev/null
+++ b/.opencode/commands/research.md
@@ -0,0 +1,30 @@
+---
+{}
+---
+
+# Research Protocol — Trigger Prompt
+#
+# Reusable prompt for research-first development.
+# MANDATORY sequence before any implementation.
+
+Research this topic before implementing: {TOPIC}
+
+## Research Sequence
+1. **Handbook search**: Search claude-context for existing patterns in /knowledge/ and docs/
+2. **Codebase search**: Search claude-context for similar implementations in this repo
+3. **Framework docs**: context7 for official documentation (Hugo, PostCSS, Ruby, etc.)
+4. **Package analysis**: package-search for dependency research (if using new packages)
+5. **Web search**: brave-search / searxng for current best practices (2025-2026)
+
+## Output Requirements
+- 5+ authoritative sources
+- Existing patterns found (or confirmation of none)
+- Framework documentation citations
+- Dependency evaluation (if applicable)
+- Recommendation with trade-offs
+
+## Anti-Patterns (BLOCKED)
+- Implementation before research
+- Single-source decisions
+- Dependencies installed without package-search verification
+- Slugs guessed without ls content/blog/ verification
\ No newline at end of file
diff --git a/.windsurf/rules/css-consolidation.md b/.windsurf/rules/css-consolidation.md
new file mode 100644
index 000000000..7094f3c5a
--- /dev/null
+++ b/.windsurf/rules/css-consolidation.md
@@ -0,0 +1,34 @@
+---
+trigger: always_on
+---
+
+# CSS Consolidation Rules — JetThoughts Site
+#
+# MANDATORY for all CSS/HTML template changes.
+# Full docs: docs/workflows/css-consolidation.md
+# Project: docs/projects/2509-css-migration/css-loading-order-analysis.md
+
+## Architecture
+- **Methodology**: BEM (Block__Element--Modifier)
+- **Pipeline**: PostCSS (import, mixins, nested, autoprefixer, cssnano, purgecss)
+- **Cascade order**: Critical → Layout → Component → Theme → Footer
+
+## Rules
+- Never modify vendor files (`css/vendors/*.min.css`) or dynamic templates (`css/dynamic-*.css`)
+- Preserve cascade order in all changes
+- Extract whole rule sets; keep page-specific `.fl-node-{hash}` in original files
+- Run `bin/rake test:critical` after each extraction
+
+## FL-Node Extraction (Beaver Builder → BEM)
+- **Phase**: TDD RED-GREEN-REFACTOR cycle
+- **Validation**: `env PRECOMPILED_ASSETS=false bin/rake test:critical` after each change
+- **Pattern**: Add BEM classes first → verify visual fidelity → remove FL-nodes → retest
+- **Commit**: micro-commit on green tests only
+
+## Visual Regression
+- Tolerance: 0.0 for refactoring (zero visual changes)
+- Tolerance: ≤0.03 for new features only
+- Run BOTH `bin/test` and `bin/dtest` before commit
+- Visual failures are commit blockers, not warnings
+
+@see docs/workflows/css-consolidation.md
diff --git a/.windsurf/rules/icp-target.md b/.windsurf/rules/icp-target.md
new file mode 100644
index 000000000..a97ceb21e
--- /dev/null
+++ b/.windsurf/rules/icp-target.md
@@ -0,0 +1,50 @@
+---
+trigger: always_on
+---
+
+# ICP Target — JetThoughts Primary Website Audience
+#
+# CANONICAL SOURCE: docs/90-99-content-strategy/strategy-analysis/90.10-icp-primary-website-target.md
+# This file is a COMPILED EXTRACT for APM — the full doc is authoritative.
+# Update the canonical source first, then sync this extract.
+#
+# MANDATORY reading before content, design, marketing, or blog work.
+# Compiled by APM into every AI tool's context.
+
+## ICP-E: Non-Technical Founder Burned by a Devshop
+
+### Who They Are
+- 35-50 years old
+- Non-technical or semi-technical
+- Raised funding ($500K–$2M seed)
+- Hired a dev shop or freelancers
+- Missed deadlines, buggy product, mounting technical debt
+- Starting to question: "Is it me? Is it them? How do I know?"
+
+### Their Pain
+- Can't evaluate code quality themselves
+- Dev shop says "almost done" for months
+- Every new feature breaks something else
+- No tests, no documentation, no deploy pipeline
+- Starting to lose investor confidence
+
+### What They Need
+- Someone who speaks plain English, not jargon
+- Proof that the problem is real and solvable (specific stories, not vague promises)
+- A clear path out: audit → stabilize → build right
+- Trust signals: named clients, specific results, honest trade-offs
+
+### What Resonates
+- Stories of other founders in the same situation
+- Specific numbers: "$40K wasted," "11-minute outage," "84 support tickets"
+- Honest acknowledgment that rescue is hard and expensive
+- The idea that good engineering is a competitive advantage
+- "You're not crazy — this IS what bad engineering looks like"
+
+### What Alienates
+- jargon-dense technical explanations before establishing trust
+- "We're the best" claims without evidence
+- Generic consulting language ("partner with you," "drive outcomes")
+- Ignoring the emotional reality (shame, frustration, fear of looking stupid)
+
+@see docs/90-99-content-strategy/strategy-analysis/90.10-icp-primary-website-target.md
diff --git a/.windsurf/rules/project-rules.md b/.windsurf/rules/project-rules.md
new file mode 100644
index 000000000..e64f154fe
--- /dev/null
+++ b/.windsurf/rules/project-rules.md
@@ -0,0 +1,83 @@
+---
+trigger: always_on
+---
+
+# JetThoughts Site — Project Rules
+#
+# This is the SINGLE SOURCE for AI agent instructions across all tools.
+# APM compiles this to each tool's expected path (CLAUDE.md, .github/
+# copilot-instructions.md, etc.). Edit here, not in tool-specific files.
+#
+# CANONICAL SOURCE: THIS FILE. The existing CLAUDE.md is a Claude-specific
+# superset; over time CLI tool instructions converge here. Voice, ICP, TDD,
+# and CSS rules are split into separate instruction files for modularity.
+#
+# Full workflow docs: docs/workflows/
+# Full content docs: docs/90-99-content-strategy/
+
+## Stack
+- **Site**: Hugo 0.147.9 Extended static site generator
+- **Theme**: Custom "Beaver" theme (BEM + PostCSS)
+- **Tests**: Ruby 3.4, Minitest, Capybara, Selenium WebDriver
+- **CSS**: PostCSS pipeline (autoprefixer, cssnano, purgecss)
+- **Package managers**: Bun (JS, NOT npm), Bundler (Ruby)
+- **CI/CD**: GitHub Actions → GitHub Pages
+
+## Build & Test Commands
+- Build: `bin/hugo-build`
+- Test (critical): `bin/rake test:critical`
+- Test (visual/macOS): `bin/test`
+- Test (visual/Linux Docker): `bin/dtest`
+- Lint Ruby: `bundle exec standardrb --fix`
+
+## Research Protocol (MANDATORY — before any change)
+1. Search codebase with `claude-context` MCP FIRST
+2. Verify exact filenames with `rg`/`ls` as fallback
+3. Read `docs/workflows/BASE_HANDBOOK.md` at session start
+4. Read `docs/workflows/flow-router.md` to route to correct workflow
+5. For blog content: find existing posts via `docs/blog-post-index.md`
+6. Never guess slugs — verify with `ls content/blog/<slug>/index.md`
+
+## Non-Negotiables
+- **Zero generic AI language** in content — all AI-sounding phrases rejected
+- **Zero Hugo build breaks** — `bin/hugo-build` must pass for content changes
+- **Zero unsupported claims** — all assertions must have citations
+- **No ad hoc test scripts** — use `bin/rake test:critical` and `bin/test`
+- **No Python for analysis** — prefer `rg`, `sed`, `awk`, shell tools
+- **No duplicate throwaway files** — no `*_refactored.*`, `*_new.*`, `*_backup.*`
+- **New docs only under `docs/workflows/`** — else edit existing files
+- **Never commit coordinator/verification reports** — write to `/tmp/` instead
+
+## Content Work (Blog Posts, LinkedIn, Pages)
+- **BEFORE writing**: Read voice guide + ICP target (see instructions/voice-guide, instructions/icp-target)
+- **Blog posts**: Follow `docs/workflows/blog-pipeline.md` end-to-end
+- **LinkedIn**: Follow `docs/workflows/linkedin-post-pipeline.md`
+- **Cover images**: Follow `docs/workflows/cover-images.md` + `.stitch/design.md`
+- Run slop-detector + humanizer pass on all drafts
+- Cold-eyes gate (9 checks) mandatory before publishing
+
+## CSS / HTML Changes
+- **BEM methodology** mandatory for all new CSS
+- **FL-node extraction**: TDD cycle — RED → GREEN → REFACTOR → TEST
+- **Visual regression tolerance**: 0.0 for refactors, ≤0.03 for new features
+- **MANDATORY**: Run BOTH `bin/test` (macOS) AND `bin/dtest` (Linux Docker) before commit
+- **Feature branches**: HTML/CSS/template changes → feature branch → one PR per sprint
+- Zero tolerance: CSS class existence testing, framework selector usage in tests
+
+## TDD & Testing
+- **Cycle**: RED → GREEN → REFACTOR
+- **Framework**: Minitest (test/system/, test/unit/)
+- **Behavior-focused tests ONLY** — reject implementation/existence/config tests
+- **Avoid fragile config assertions** — assert shape, not tunable values
+- **Visual regression**: screenshot comparison, 0.0 tolerance for refactoring
+- Run `bin/rake test:critical` after every micro-change (<10 lines)
+- Chrome DevTools validation after HTML/CSS/JS changes: zero console errors, zero 404s
+
+## Behavioral Constraints
+- **User says "code is bad" or "over-engineered"**: HALT, 5-Why analysis, fix config, proceed
+- **Reflection triggers**: zero-defect tracking, test masking detection
+- **Four-eyes principle**: paired agents for all critical changes
+- **Direct-to-master only for**: content-only blog edits, commit fixes, docs, CLAUDE.md policy updates
+
+@see docs/workflows/ for full workflow documentation
+@see docs/90-99-content-strategy/ for content strategy
diff --git a/.windsurf/rules/tdd-enforcement.md b/.windsurf/rules/tdd-enforcement.md
new file mode 100644
index 000000000..0be2a001a
--- /dev/null
+++ b/.windsurf/rules/tdd-enforcement.md
@@ -0,0 +1,66 @@
+---
+trigger: always_on
+---
+
+# TDD Enforcement — JetThoughts Site
+#
+# MANDATORY for all code changes. Extracted from CLAUDE.md TDD section.
+# Full docs: docs/60.03-tdd-quality-enforcement.md, docs/60.04-four-eyes-principle.md
+
+## TDD Cycle
+1. **RED**: Write a failing test that validates BEHAVIOR (not implementation)
+2. **GREEN**: Minimal code to make the test pass
+3. **REFACTOR**: Clean up while tests stay green
+4. **COMMIT**: Micro-commit on green tests
+
+## Test Quality Rules (ZERO TOLERANCE)
+
+### Prohibited Test Patterns — Agent MUST Reject
+- ❌ CSS class existence testing (`.fl-node-*`, `.has_css?(".px-4")`)
+- ❌ HTML structure counting (`page.all("div").count`)
+- ❌ Framework-generated selector usage (FL-Builder, Tailwind utility classes)
+- ❌ Method existence testing (`assert SomeClass.method_defined?`)
+- ❌ Config value assertions (hardcoded `q=90`, exact file sizes, specific dimensions)
+- ❌ Implementation detail testing
+
+### Required Test Patterns
+- ✅ Visual regression: `take_screenshot` → `assert_visual_match` with tolerance
+- ✅ User-visible content: `assert page.has_content?("text")`
+- ✅ Semantic elements: `assert page.has_css?("h1", text: "...")`
+- ✅ User interaction: `click_button` → `assert_current_path`
+- ✅ Shape assertions: `q=\d+`, `src contains wsrv.nl`, not `q=90`, not exact URLs
+
+### Avoid Fragile Config Assertions
+Don't hardcode tunable values. Assert the *shape*, not the configuration:
+- Bad: `assert_match /q=90/, src`
+- Good: `assert_match /q=\d+/, src`
+- Bad: `assert_match /w=360/, src`
+- Good: `assert has_css?("picture source")`
+
+## Visual Regression Gates (BLOCKING for HTML/CSS changes)
+- **Tolerance**: 0.0 for refactoring, ≤0.03 for new features
+- **Protocol**: Capture baseline BEFORE changes → compare AFTER → block on any diff >0%
+- **MANDATORY**: Run BOTH `bin/test` (macOS) AND `bin/dtest` (Linux Docker) before commit
+- Visual regression failures are COMMIT BLOCKERS, not warnings
+
+## Test Runner
+- Framework: Minitest
+- Critical tests: `bin/rake test:critical` (after every micro-change <10 lines)
+- Visual macOS: `bin/test`
+- Visual Linux/Docker: `bin/dtest`
+- NEVER create ad hoc `*.sh` test scripts
+
+## Chrome DevTools Validation (MANDATORY after HTML/CSS/JS)
+1. Start Hugo dev server, open page in Chrome DevTools
+2. Console: ZERO JS errors
+3. Network: ZERO 404s
+4. Capture desktop + mobile screenshots
+5. All gates pass → commit. Any gate fails → fix, repeat
+
+## Feature Branch + PR (HTML/CSS/template changes)
+1. `git checkout -b <sprint-name>`
+2. Multiple related commits on branch
+3. Both test gates green on each commit
+4. `git push -u origin <branch>`
+5. ONE PR per sprint via `gh pr create`
+6. Direct-to-master only for: content-only blog edits, commit fixes, docs, CLAUDE.md updates
diff --git a/.windsurf/rules/voice-guide.md b/.windsurf/rules/voice-guide.md
new file mode 100644
index 000000000..ed72a3fac
--- /dev/null
+++ b/.windsurf/rules/voice-guide.md
@@ -0,0 +1,72 @@
+---
+trigger: always_on
+---
+
+# Voice Guide — JetThoughts Content
+#
+# CANONICAL SOURCE: docs/90-99-content-strategy/strategy-analysis/90.11-voice-guide.md
+# This file is a COMPILED EXTRACT for APM — the full doc is authoritative.
+# Update the canonical source first, then sync this extract.
+#
+# MANDATORY reading before writing ANY blog post, LinkedIn post, or marketing copy.
+# Compiled by APM into every AI tool's context.
+
+## Two Voices
+
+### Founder Posts (Paul Keen / JT leadership)
+- **Rob Walling directness** + **Rand Fishkin vulnerability**
+- Plain English, empathy, specific numbers
+- Talk to a non-technical founder burned by a devshop
+- "Been there" energy — practitioner not vendor
+- Trade-offs acknowledged honestly
+
+### Tech Posts (engineering team)
+- **Thoughtbot style**: open with tension, not features
+- Own your opinions — no "it depends" hedging
+- Code as narrative evidence, not decoration
+- Honest trade-offs: what does this solution fail to do?
+
+## Banned Words
+unlock, harness, leverage, dive in, game-changer, journey, seamless, robust,
+supercharge, revolutionize, embark, delve, discover, blast radius, suspect list
+
+## Banned Structural Patterns
+- Rule of three (parallel triads)
+- Signposting ("This post is about...", "Here's what you need to know")
+- Bold inline-header lists
+- Triple rhetorical questions
+- Slogany parallel closings ("X is A. Y is B.")
+- Therapist voice ("That's the hardest part")
+- Copula avoidance ("serves as" → use "is")
+- Impersonal fragment stacking — every sentence needs a human subject
+- Sustained staccato (3+ consecutive short fragments or uniform-length sentences)
+- Command structure repetition (3+ paragraphs starting with imperatives)
+- Apologetic caveats ("every project is different")
+- Telling instead of showing — replace adjectives with concrete mechanics
+- Fake authority generalizations ("We've seen this on every codebase")
+- Timeline fabrication — verify technology release dates before using "last year"
+- Use `-` not `—` for all dashes
+
+## Quality Tests (run on every paragraph)
+- **"Who" test**: who is doing what? If subject is a thing, rewrite with a person
+- **"Show" test**: did I show the mechanic or just label it?
+- **"Coffee" test**: would a tired founder talk like this at a bar?
+- **"90/10" test**: is ≥90% educational? Is promotion ≤10%?
+- **"Trade-off" test**: did I acknowledge what the solution fails to do?
+- **"Practitioner" test**: replace generalized scenarios with specific incidents
+
+## Mannerism Noun Cap
+discipline, cycle, pattern, canonical, workflow, framework, cadence, rhythm,
+contract, gate, gatekeeper, layer, round, review, pass
+→ Max 5 occurrences total per post. Vary to the concrete thing happening.
+
+## Cadence Quotas (per H2 section)
+- At least one sentence <9 words AND one >22 words
+- No 3 consecutive sentences within 3 words of each other
+- One single-sentence paragraph per H2
+- Max 3 sentences / 70 words per paragraph
+- No two sentences in same paragraph starting with same first word
+- "The..." paragraph openers: <20% of total paragraphs
+- "We..." paragraph openers: <15% of total paragraphs
+
+@see docs/90-99-content-strategy/strategy-analysis/90.11-voice-guide.md
diff --git a/.windsurf/workflows/blog-pipeline.md b/.windsurf/workflows/blog-pipeline.md
new file mode 100644
index 000000000..5c9dae536
--- /dev/null
+++ b/.windsurf/workflows/blog-pipeline.md
@@ -0,0 +1,79 @@
+---
+{}
+---
+
+# Blog Post Pipeline — Trigger Prompt
+#
+# Copy this prompt to write the next blog post from the content plan.
+# Full pipeline: docs/workflows/blog-pipeline.md
+#
+# This is a reusable PROMPT — APM compiles it to slash-commands for
+# each AI tool (e.g., /blog-post in Claude, #blog-post in Copilot).
+
+Write the next blog post following the full pipeline.
+
+## STEP 1 — Pick Topic
+Read the content plan: docs/projects/2510-seo-content-strategy/20-29-strategy/20.07-content-plan-icp-e-q2-2026.md
+Find the next 🔲 Planned topic. Confirm slug, primary keyword, target publish date.
+
+## STEP 2 — Read Context (mandatory, do not skip)
+- ICP: docs/90-99-content-strategy/strategy-analysis/90.10-icp-primary-website-target.md
+- Voice guide: docs/90-99-content-strategy/strategy-analysis/90.11-voice-guide.md
+- Content plan brief: read the topic's brief section
+- Design spec: .stitch/design.md
+
+## STEP 3 — Research
+- Web search for 5+ authoritative sources
+- Check existing posts for duplication: ls content/blog/ | grep -iE "<key-terms>"
+- Verify every statistic has a citable source
+
+## STEP 3b — Find Internal Posts to Reference
+- Use claude-context MCP search FIRST
+- For tag/slug lookups: docs/blog-post-index.md
+- Verify each slug: ls content/blog/<slug>/index.md
+- Collect at least 4 verified slugs before drafting
+
+## STEP 4 — Draft
+Create content/blog/<slug>/index.md with:
+- Frontmatter: title ≤45 chars, description 150-160 chars, date, slug, keywords, tags
+- Voice: plain English, empathy first, founder-to-founder tone
+- Structure: punchy hook → keyword in first 100 words → numbered H2s → "When NOT to" → JT proof → CTA → related reading → further reading
+- 4+ internal links in body, 5+ external citations
+- Paragraphs ≤3 sentences
+- Cut test: "what does the reader DO differently?" — delete if answer is "nothing new"
+
+## STEP 5 — Review Loop
+Spawn 3 parallel critic agents:
+1. Founder persona (Alex, burned CEO) — ≥8/10
+2. SEO/slop detector — ≥8/10 humanization
+3. Senior copy editor — "publish as-is" or "minor edits"
+
+## STEP 5c — Cold-Eyes Gate (mandatory)
+Spawn fresh-context reviewer with 9 explicit checks:
+1. Frontmatter vs body consistency
+2. Fluffy AI narration detection
+3. Fake authority stamps
+4. Timeline fabrication
+5. "Not X but Y" patterns
+6. Em dashes → hyphens
+7. Rule-of-three modifiers
+8. Cross-section template repetition
+9. Impersonal fragment stacking
+
+## STEP 6 — Cover Image
+- Read .stitch/design.md (6-slot layout)
+- Duplicate existing HTML from .stitch/designs/
+- Render: chrome-devtools → viewport 2400×1260×2 → screenshot
+- Downsample: magick → 2400×1260 → content/blog/<slug>/cover.png
+
+## STEP 7 — Validate
+- bin/hugo-build must pass (zero errors)
+- Chrome DevTools: zero console errors, zero 404s, og:image resolves
+
+## STEP 8 — Update Content Plan
+- Change topic status: 🔲 Planned → ✅ Published
+- Add commit SHA + changelog
+
+## STEP 9 — Commit
+- Stage: post directory + cover HTML + content plan update
+- Commit format: feat(content): add <topic> post
\ No newline at end of file
diff --git a/.windsurf/workflows/cover-image.md b/.windsurf/workflows/cover-image.md
new file mode 100644
index 000000000..024cd8573
--- /dev/null
+++ b/.windsurf/workflows/cover-image.md
@@ -0,0 +1,23 @@
+---
+{}
+---
+
+# Cover Image Generation — Trigger Prompt
+#
+# Reusable prompt for generating blog post cover images via Stitch design system.
+# Full spec: .stitch/design.md (canonical 6-slot layout)
+# Full workflow: docs/workflows/cover-images.md
+
+Generate a cover image for a blog post.
+
+## Steps
+1. Read .stitch/design.md for the 6-slot layout specification
+2. Duplicate an existing HTML from .stitch/designs/ (e.g., rails-8-1-active-job-continuations-cover.html)
+3. Change ONLY: title, category, pill, 3 headline lines, 3 chip labels+values, status text
+4. Render: chrome-devtools new_page → emulate viewport=2400x1260x2 → wait 2s → take_screenshot
+5. Downsample: magick {screenshot} -filter Lanczos -resize 2400x1260 -sampling-factor 4:4:4 -strip -quality 95 content/blog/<slug>/cover.png
+
+## Design System
+- JetVelocity: obsidian dark, Ruby red (#cc342d), neon purple (#a855f7)
+- Font: Inter (headlines), JetBrains Mono (code/metrics)
+- 6-slot grid layout — see .stitch/design.md for exact positioning
\ No newline at end of file
diff --git a/.windsurf/workflows/research.md b/.windsurf/workflows/research.md
new file mode 100644
index 000000000..59101394b
--- /dev/null
+++ b/.windsurf/workflows/research.md
@@ -0,0 +1,30 @@
+---
+{}
+---
+
+# Research Protocol — Trigger Prompt
+#
+# Reusable prompt for research-first development.
+# MANDATORY sequence before any implementation.
+
+Research this topic before implementing: {TOPIC}
+
+## Research Sequence
+1. **Handbook search**: Search claude-context for existing patterns in /knowledge/ and docs/
+2. **Codebase search**: Search claude-context for similar implementations in this repo
+3. **Framework docs**: context7 for official documentation (Hugo, PostCSS, Ruby, etc.)
+4. **Package analysis**: package-search for dependency research (if using new packages)
+5. **Web search**: brave-search / searxng for current best practices (2025-2026)
+
+## Output Requirements
+- 5+ authoritative sources
+- Existing patterns found (or confirmation of none)
+- Framework documentation citations
+- Dependency evaluation (if applicable)
+- Recommendation with trade-offs
+
+## Anti-Patterns (BLOCKED)
+- Implementation before research
+- Single-source decisions
+- Dependencies installed without package-search verification
+- Slugs guessed without ls content/blog/ verification
\ No newline at end of file
diff --git a/AGENTS.md b/AGENTS.md
index 89de46797..d6d13dffc 100644
--- a/AGENTS.md
+++ b/AGENTS.md
@@ -1,27 +1,16 @@
 # AGENTS.md — JetThoughts Blog
+#
+# GENERATED BY APM — do not edit directly.
+# Source: .apm/instructions/project-rules.instructions.md (manual — no APM target for AGENTS.md)
+# Regenerate: apm install
 
-Canonical instructions live in `CLAUDE.md`. Read that first.
-Session start: always read `@docs/workflows/BASE_HANDBOOK.md` and `@docs/workflows/flow-router.md`.
+Canonical instructions: `.apm/instructions/project-rules.instructions.md`
+Canonical voice guide: `.apm/instructions/voice-guide.instructions.md`
+Canonical ICP target: `.apm/instructions/icp-target.instructions.md`
 
-**ICP (MANDATORY for content/design/marketing work):** Read `docs/90-99-content-strategy/strategy-analysis/90.10-icp-primary-website-target.md` before writing blog posts, designing pages, planning content, or creating marketing materials. Target audience: non-technical founder burned by a devshop.
+Session start: read `docs/workflows/BASE_HANDBOOK.md` and `docs/workflows/flow-router.md`.
 
-**Voice (MANDATORY when writing):** Read `docs/90-99-content-strategy/strategy-analysis/90.11-voice-guide.md`.
-- **Founder posts**: Rob Walling directness + Rand Fishkin vulnerability. Plain English, empathy, specific numbers.
-- **Tech posts**: Thoughtbot style. Open with tension not features. Own opinions. Code as narrative evidence. Honest tradeoffs.
-- Run the stream-specific self-test checklist before publishing.
-
-**Finding posts to reference (MANDATORY when writing blog posts):** Use claude-context MCP search first: `Search the codebase at /Users/pftg/dev/jetthoughts.github.io for: "topic keywords"`. For tag/slug lookups see `docs/blog-post-index.md`. Never guess slugs — verify with `ls content/blog/<slug>/index.md`.
-
-Workflow references:
-- `docs/workflows/css-consolidation.md`
-- `docs/workflows/blog-pipeline.md`
-- `docs/workflows/cover-images.md` (canonical spec in `.stitch/design.md`)
-- `docs/workflows/commands.md`
-- `docs/workflows/agents.md`
-
-Key commands:
-- `bin/hugo-build`
-- `bin/rake test:critical`
+Key commands: `bin/hugo-build` | `bin/rake test:critical`
 
 <!-- lean-ctx-compression -->
 OUTPUT STYLE: expert-terse
diff --git a/Brewfile b/Brewfile
index 91244919d..eca2a1588 100644
--- a/Brewfile
+++ b/Brewfile
@@ -4,3 +4,6 @@ brew "bun", link: true, force: true
 brew "hugo"
 brew "rbenv"
 brew "ruby-build"
+
+# MCP infrastructure
+brew "ollama", restart_service: true
diff --git a/GEMINI.md b/GEMINI.md
index efaf014bd..443d4375e 100644
--- a/GEMINI.md
+++ b/GEMINI.md
@@ -1,21 +1,13 @@
 # GEMINI.md — JetThoughts Blog
+#
+# GENERATED BY APM — do not edit directly.
+# Source: apm.yml → targets.gemini → .apm/instructions/project-rules.instructions.md
+# Regenerate: apm install
 
-Canonical instructions live in `CLAUDE.md`. Read that first.
-Session start: always read `@docs/workflows/BASE_HANDBOOK.md` and `@docs/workflows/flow-router.md`.
+Canonical instructions: `.apm/instructions/project-rules.instructions.md`
+Canonical voice guide: `.apm/instructions/voice-guide.instructions.md`
+Canonical ICP target: `.apm/instructions/icp-target.instructions.md`
 
-**ICP (MANDATORY for content/design/marketing):** Read `docs/90-99-content-strategy/strategy-analysis/90.10-icp-primary-website-target.md` before any content or design work. Target: non-technical founder burned by a devshop.
+Session start: read `docs/workflows/BASE_HANDBOOK.md` and `docs/workflows/flow-router.md`.
 
-**Voice (MANDATORY when writing):** Read `docs/90-99-content-strategy/strategy-analysis/90.11-voice-guide.md`. Founder posts: Walling+Fishkin. Tech posts: thoughtbot style (tension-first, opinionated, code as evidence). Run stream-specific self-test.
-
-**Finding posts to reference:** Use claude-context MCP search: `Search the codebase at /Users/pftg/dev/jetthoughts.github.io for: "topic"`. Tag/slug index: `docs/blog-post-index.md`. Verify slugs: `ls content/blog/<slug>/index.md`.
-
-Workflow references:
-- `docs/workflows/css-consolidation.md`
-- `docs/workflows/blog-pipeline.md`
-- `docs/workflows/cover-images.md` (canonical spec in `.stitch/design.md`)
-- `docs/workflows/commands.md`
-- `docs/workflows/agents.md`
-
-Key commands:
-- `bin/hugo-build`
-- `bin/rake test:critical`
+Key commands: `bin/hugo-build` | `bin/rake test:critical`
diff --git a/Makefile b/Makefile
new file mode 100644
index 000000000..2713bb30c
--- /dev/null
+++ b/Makefile
@@ -0,0 +1,52 @@
+# Makefile — JetThoughts Site
+#
+# APM declares MCP dependencies in apm.yml (what agents need).
+# This Makefile RUNS the persistent backends via docker compose + brew.
+# Stdio-based MCPs are auto-started by the AI tool — no target needed.
+#
+# Quick start:
+#   apm install      # Install agent primitives (instructions, agents, prompts)
+#   make mcp-up      # Start persistent MCP backends
+#   make mcp-status  # Verify everything is running
+#   code .           # Open in Claude/Copilot/Cursor/etc.
+
+.PHONY: mcp-up mcp-down mcp-status help
+
+help:
+	@echo "JetThoughts Site — Development Commands"
+	@echo ""
+	@echo "  make mcp-up       Start all MCP backends (ollama + docker compose)"
+	@echo "  make mcp-down     Stop Docker-based backends"
+	@echo "  make mcp-status   Health check all MCP backends"
+	@echo ""
+	@echo "Docker services defined in compose.mcp.yml"
+	@echo "Ollama managed via brew services"
+	@echo "MCP servers declared in apm.yml → dependencies.mcp"
+
+mcp-up:
+	@echo "==> Starting MCP backends..."
+	@echo ""
+	# Ollama — local LLM/embeddings via Homebrew
+	@brew services start ollama 2>/dev/null && echo "   ✓ ollama started" || echo "   ! ollama not installed — run: brew install ollama"
+	@echo ""
+	# Docker-based MCP services (searxng, milvus, etcd, minio)
+	@docker compose -f compose.mcp.yml up -d
+	@echo ""
+	@$(MAKE) mcp-status
+
+mcp-down:
+	@echo "==> Stopping MCP backends..."
+	@docker compose -f compose.mcp.yml down
+	@echo "   ollama: left running (brew services stop ollama to stop)"
+	@echo "==> MCP backends stopped"
+
+mcp-status:
+	@echo "==> MCP Backend Status"
+	@echo ""
+	@printf "   %-14s " "ollama :11434"; \
+		curl -s -o /dev/null -w "%{http_code}" http://localhost:11434/api/tags 2>/dev/null | grep -q 200 && echo "✓ running" || echo "✗ down"
+	@printf "   %-14s " "searxng :8080"; \
+		curl -s -o /dev/null -w "%{http_code}" http://localhost:8080/healthz 2>/dev/null | grep -q 200 && echo "✓ running" || echo "✗ down"
+	@printf "   %-14s " "milvus :19530"; \
+		curl -s -o /dev/null -w "%{http_code}" http://localhost:19530/health 2>/dev/null | grep -q 200 && echo "✓ running" || echo "✗ down"
+	@echo ""
diff --git a/QWEN.md b/QWEN.md
index 10b0eb6d4..05bf21c8f 100644
--- a/QWEN.md
+++ b/QWEN.md
@@ -1,24 +1,14 @@
 # QWEN.md — JetThoughts Blog
+#
+# GENERATED BY APM — do not edit directly.
+# Source: .apm/instructions/project-rules.instructions.md (manual — Qwen Code not yet an APM target)
+# Regenerate: apm install
 
-Canonical instructions live in `CLAUDE.md`. Read that first.
+Canonical instructions: `.apm/instructions/project-rules.instructions.md`
+Canonical voice guide: `.apm/instructions/voice-guide.instructions.md`
+Canonical ICP target: `.apm/instructions/icp-target.instructions.md`
 
-Session start: always read `@docs/workflows/BASE_HANDBOOK.md` and `@docs/workflows/flow-router.md`.
+Session start: read `docs/workflows/BASE_HANDBOOK.md` and `docs/workflows/flow-router.md`.
+Prefer **skills** over agents. Use agents only when explicitly required.
 
-**ICP (MANDATORY for content/design/marketing):** Read `docs/90-99-content-strategy/strategy-analysis/90.10-icp-primary-website-target.md` before any content or design work. Target: non-technical founder burned by a devshop.
-
-**Voice (MANDATORY when writing):** Read `docs/90-99-content-strategy/strategy-analysis/90.11-voice-guide.md`. Founder posts: Walling+Fishkin. Tech posts: thoughtbot style (tension-first, opinionated, code as evidence). Run stream-specific self-test.
-
-**Finding posts to reference:** Use claude-context MCP search: `Search the codebase at /Users/pftg/dev/jetthoughts.github.io for: "topic"`. Tag/slug index: `docs/blog-post-index.md`. Verify slugs: `ls content/blog/<slug>/index.md`.
-
-Prefer **skills** over agents. Use agents only when a task explicitly requires them.
-
-Workflow references:
-- `docs/workflows/css-consolidation.md`
-- `docs/workflows/blog-pipeline.md`
-- `docs/workflows/cover-images.md` (canonical spec in `.stitch/design.md`)
-- `docs/workflows/commands.md`
-- `docs/workflows/agents.md`
-
-Key commands:
-- `bin/hugo-build`
-- `bin/rake test:critical`
+Key commands: `bin/hugo-build` | `bin/rake test:critical`
diff --git a/README.md b/README.md
index 16997157d..3f1c0fea2 100644
--- a/README.md
+++ b/README.md
@@ -12,6 +12,7 @@
 - [Features](#features)
 - [Technology Stack](#technology-stack)
 - [Quick Start](#quick-start)
+- [AI Agent Setup (APM)](#ai-agent-setup-apm)
 - [Development](#development)
 - [Testing](#testing)
 - [Deployment](#deployment)
@@ -130,6 +131,60 @@ hugo server -D
 
 Visit `http://localhost:1313` to view the site.
 
+## 🤖 AI Agent Setup (APM)
+
+This project uses [Microsoft APM](https://github.com/microsoft/apm) (Agent Package Manager) to manage AI coding tool configuration. One manifest (`apm.yml`) declares instructions, agents, skills, prompts, and MCP servers — and APM compiles them for **all** tools automatically.
+
+### Tools Auto-Configured
+
+Claude Code • GitHub Copilot • Codex CLI • Cursor • Gemini CLI • Windsurf • Devin • Codebuff • Qwen Code • Junie
+
+### Quick Setup
+
+```bash
+# 1. Install APM (macOS)
+curl -fsSL https://get.apm.dev | sh
+# or via Homebrew: brew install microsoft/tap/apm
+
+# 2. Install agent primitives (instructions, agents, prompts)
+apm install
+
+# 3. Start persistent MCP backends (ollama, searxng, milvus)
+make mcp-up
+
+# 4. Open in any AI tool — same config everywhere
+code .                    # VS Code with Copilot
+claude                    # Claude Code
+gemini                    # Gemini CLI
+```
+
+### What APM Manages
+
+| Primitive | Location | Purpose |
+|-----------|----------|---------|
+| **Instructions** | `.apm/instructions/` | Shared project rules compiled to all tools |
+| **Prompts** | `.apm/prompts/` | Slash-command triggers (blog-pipeline, research, cover-image) |
+| **Agents** | `.apm/agents/` | Persona definitions (coder, reviewer, content-creator, etc.) |
+| **Skills** | `.apm/skills/` | Reusable capabilities (blog-pipeline, css-bem-migration, etc.) |
+| **MCP Servers** | Declared in `apm.yml` | claude-context, context7, brave-search, searxng, ollama, milvus, etc. |
+
+### MCP Servers
+
+MCP servers are declared in `apm.yml` → `dependencies.mcp`. Two categories:
+
+- **Stdio (auto-started)**: `claude-context`, `context7`, `brave-search`, `lean-ctx`, `chrome-devtools`, `package-search`, `github-mcp-server` — started on-demand by the AI tool.
+- **Persistent**: `ollama` (brew service), `searxng` + `milvus` (docker compose via `compose.mcp.yml`) — managed via `make mcp-up` / `make mcp-down`.
+
+### Environment Variables
+
+API keys for MCP servers are consumed from your environment — never hardcoded in config files:
+
+```bash
+# MCP Server API Keys
+BRAVE_API_KEY=your_brave_search_key
+GITHUB_TOKEN=your_github_pat
+```
+
 ## 💻 Development
 
 ### Project Structure
diff --git a/apm.lock.yaml b/apm.lock.yaml
new file mode 100644
index 000000000..a78be1249
--- /dev/null
+++ b/apm.lock.yaml
@@ -0,0 +1,395 @@
+lockfile_version: '1'
+generated_at: '2026-07-08T02:51:37.719380+00:00'
+dependencies: []
+mcp_servers:
+- brave-search
+- chrome-devtools
+- claude-context
+- context7
+- github-mcp-server
+- lean-ctx
+- milvus
+- ollama
+- package-search
+- searxng
+mcp_configs:
+  brave-search:
+    name: brave-search
+    transport: stdio
+    env:
+      BRAVE_API_KEY: ${BRAVE_API_KEY}
+    args:
+    - -y
+    - '@anthropic/brave-search-mcp'
+    registry: false
+    command: npx
+  chrome-devtools:
+    name: chrome-devtools
+    transport: stdio
+    args:
+    - -y
+    - '@anthropic/chrome-devtools-mcp'
+    registry: false
+    command: npx
+  claude-context:
+    name: claude-context
+    transport: stdio
+    args:
+    - -y
+    - '@anthropic/claude-context'
+    registry: false
+    command: npx
+  context7:
+    name: context7
+    transport: stdio
+    args:
+    - -y
+    - '@upstash/context7-mcp'
+    registry: false
+    command: npx
+  github-mcp-server:
+    name: github-mcp-server
+    transport: stdio
+    env:
+      GITHUB_TOKEN: ${GITHUB_TOKEN}
+    args:
+    - -y
+    - '@anthropic/github-mcp-server'
+    registry: false
+    command: npx
+  lean-ctx:
+    name: lean-ctx
+    transport: stdio
+    args:
+    - -y
+    - lean-ctx-mcp
+    registry: false
+    command: npx
+  milvus:
+    name: milvus
+    transport: http
+    registry: false
+    url: http://localhost:19530
+  ollama:
+    name: ollama
+    transport: http
+    registry: false
+    url: http://localhost:11434
+  package-search:
+    name: package-search
+    transport: stdio
+    args:
+    - -y
+    - package-search-mcp
+    registry: false
+    command: npx
+  searxng:
+    name: searxng
+    transport: http
+    registry: false
+    url: http://localhost:8080
+local_deployed_files:
+- .agents/skills/blog-post-pipeline
+- .agents/skills/blog-post-pipeline/SKILL.md
+- .agents/skills/content-voice-enforcement
+- .agents/skills/content-voice-enforcement/SKILL.md
+- .agents/skills/cover-image-generation
+- .agents/skills/cover-image-generation/SKILL.md
+- .agents/skills/css-bem-migration
+- .agents/skills/css-bem-migration/SKILL.md
+- .agents/skills/visual-regression-testing
+- .agents/skills/visual-regression-testing/SKILL.md
+- .claude/agents/analyst.md
+- .claude/agents/coder.md
+- .claude/agents/content-marketer.md
+- .claude/agents/planner.md
+- .claude/agents/qa-browser-tester.md
+- .claude/agents/researcher.md
+- .claude/agents/reviewer.md
+- .claude/agents/ruby-coder.md
+- .claude/agents/ruby-reviewer.md
+- .claude/agents/screenshot-guardian.md
+- .claude/agents/tester.md
+- .claude/agents/tutorial-engineer.md
+- .claude/agents/ui-problem-diagnosis.md
+- .claude/agents/ux-browser-validator.md
+- .claude/commands/blog-pipeline.md
+- .claude/commands/cover-image.md
+- .claude/commands/research.md
+- .claude/rules/css-consolidation.md
+- .claude/rules/icp-target.md
+- .claude/rules/project-rules.md
+- .claude/rules/tdd-enforcement.md
+- .claude/rules/voice-guide.md
+- .claude/skills/blog-post-pipeline
+- .claude/skills/blog-post-pipeline/SKILL.md
+- .claude/skills/content-voice-enforcement
+- .claude/skills/content-voice-enforcement/SKILL.md
+- .claude/skills/cover-image-generation
+- .claude/skills/cover-image-generation/SKILL.md
+- .claude/skills/css-bem-migration
+- .claude/skills/css-bem-migration/SKILL.md
+- .claude/skills/visual-regression-testing
+- .claude/skills/visual-regression-testing/SKILL.md
+- .codex/agents/analyst.toml
+- .codex/agents/build-monitor.toml
+- .codex/agents/coder.toml
+- .codex/agents/content-creator.toml
+- .codex/agents/content-marketer.toml
+- .codex/agents/hugo-expert.toml
+- .codex/agents/planner.toml
+- .codex/agents/qa-browser-tester.toml
+- .codex/agents/researcher.toml
+- .codex/agents/reviewer.toml
+- .codex/agents/ruby-coder.toml
+- .codex/agents/ruby-reviewer.toml
+- .codex/agents/screenshot-guardian.toml
+- .codex/agents/seo-specialist.toml
+- .codex/agents/tester.toml
+- .codex/agents/tutorial-engineer.toml
+- .codex/agents/ui-problem-diagnosis.toml
+- .codex/agents/ux-browser-validator.toml
+- .cursor/agents/analyst.md
+- .cursor/agents/build-monitor.md
+- .cursor/agents/coder.md
+- .cursor/agents/content-creator.md
+- .cursor/agents/content-marketer.md
+- .cursor/agents/hugo-expert.md
+- .cursor/agents/planner.md
+- .cursor/agents/qa-browser-tester.md
+- .cursor/agents/researcher.md
+- .cursor/agents/reviewer.md
+- .cursor/agents/ruby-coder.md
+- .cursor/agents/ruby-reviewer.md
+- .cursor/agents/screenshot-guardian.md
+- .cursor/agents/seo-specialist.md
+- .cursor/agents/tester.md
+- .cursor/agents/tutorial-engineer.md
+- .cursor/agents/ui-problem-diagnosis.md
+- .cursor/agents/ux-browser-validator.md
+- .cursor/commands/blog-pipeline.md
+- .cursor/commands/cover-image.md
+- .cursor/commands/research.md
+- .cursor/rules/css-consolidation.mdc
+- .cursor/rules/icp-target.mdc
+- .cursor/rules/project-rules.mdc
+- .cursor/rules/tdd-enforcement.mdc
+- .cursor/rules/voice-guide.mdc
+- .gemini/commands/blog-pipeline.toml
+- .gemini/commands/cover-image.toml
+- .gemini/commands/research.toml
+- .github/agents/analyst.agent.md
+- .github/agents/build-monitor.agent.md
+- .github/agents/coder.agent.md
+- .github/agents/content-creator.agent.md
+- .github/agents/content-marketer.agent.md
+- .github/agents/hugo-expert.agent.md
+- .github/agents/planner.agent.md
+- .github/agents/qa-browser-tester.agent.md
+- .github/agents/researcher.agent.md
+- .github/agents/reviewer.agent.md
+- .github/agents/ruby-coder.agent.md
+- .github/agents/ruby-reviewer.agent.md
+- .github/agents/screenshot-guardian.agent.md
+- .github/agents/seo-specialist.agent.md
+- .github/agents/tester.agent.md
+- .github/agents/tutorial-engineer.agent.md
+- .github/agents/ui-problem-diagnosis.agent.md
+- .github/agents/ux-browser-validator.agent.md
+- .github/instructions/css-consolidation.instructions.md
+- .github/instructions/icp-target.instructions.md
+- .github/instructions/project-rules.instructions.md
+- .github/instructions/tdd-enforcement.instructions.md
+- .github/instructions/voice-guide.instructions.md
+- .github/prompts/blog-pipeline.prompt.md
+- .github/prompts/cover-image.prompt.md
+- .github/prompts/research.prompt.md
+- .kiro/skills/blog-post-pipeline
+- .kiro/skills/blog-post-pipeline/SKILL.md
+- .kiro/skills/content-voice-enforcement
+- .kiro/skills/content-voice-enforcement/SKILL.md
+- .kiro/skills/cover-image-generation
+- .kiro/skills/cover-image-generation/SKILL.md
+- .kiro/skills/css-bem-migration
+- .kiro/skills/css-bem-migration/SKILL.md
+- .kiro/skills/visual-regression-testing
+- .kiro/skills/visual-regression-testing/SKILL.md
+- .kiro/steering/css-consolidation.md
+- .kiro/steering/icp-target.md
+- .kiro/steering/project-rules.md
+- .kiro/steering/tdd-enforcement.md
+- .kiro/steering/voice-guide.md
+- .opencode/agents/analyst.md
+- .opencode/agents/build-monitor.md
+- .opencode/agents/coder.md
+- .opencode/agents/content-creator.md
+- .opencode/agents/content-marketer.md
+- .opencode/agents/hugo-expert.md
+- .opencode/agents/planner.md
+- .opencode/agents/qa-browser-tester.md
+- .opencode/agents/researcher.md
+- .opencode/agents/reviewer.md
+- .opencode/agents/ruby-coder.md
+- .opencode/agents/ruby-reviewer.md
+- .opencode/agents/screenshot-guardian.md
+- .opencode/agents/seo-specialist.md
+- .opencode/agents/tester.md
+- .opencode/agents/tutorial-engineer.md
+- .opencode/agents/ui-problem-diagnosis.md
+- .opencode/agents/ux-browser-validator.md
+- .opencode/commands/blog-pipeline.md
+- .opencode/commands/cover-image.md
+- .opencode/commands/research.md
+- .windsurf/rules/css-consolidation.md
+- .windsurf/rules/icp-target.md
+- .windsurf/rules/project-rules.md
+- .windsurf/rules/tdd-enforcement.md
+- .windsurf/rules/voice-guide.md
+- .windsurf/workflows/blog-pipeline.md
+- .windsurf/workflows/cover-image.md
+- .windsurf/workflows/research.md
+local_deployed_file_hashes:
+  .agents/skills/blog-post-pipeline/SKILL.md: sha256:f169ad41677c3e50d367f8edbd3b765d8f853dd967a70201e85e79ffe4b31356
+  .agents/skills/content-voice-enforcement/SKILL.md: sha256:0005acf787979ed0935833437e5225fc893a4c8ee5b206e19f81df0ebafb56f6
+  .agents/skills/cover-image-generation/SKILL.md: sha256:1b386045354b231a9575d682e9539e4244362ba686877f49985f1914002a1314
+  .agents/skills/css-bem-migration/SKILL.md: sha256:cfd51b99fefc745d40c5fa4cb7ea1755cea6b5d1f9ee4ea819c0cfd78f9fb85c
+  .agents/skills/visual-regression-testing/SKILL.md: sha256:952575d548ed20abce7e08f79bec4ba53a4c72987de0b507520069d332f8e10b
+  .claude/agents/analyst.md: sha256:fd6b3b5c78aec9a2521e6e39f4256598e9aa0d01dedcbb2357a44cf5d4caf3e6
+  .claude/agents/coder.md: sha256:60533575db22019d7a1a513e6efe854c5b59d1859f146d38132b123ffbb4aa17
+  .claude/agents/content-marketer.md: sha256:1cd7ee611816579e75a1c6c2ea8ce9a20a4712010da5b59848913a08b4d43ffe
+  .claude/agents/planner.md: sha256:7dd9e396043bdd52989698f25e47c9cd1dc75b901677388ab062ff3237737706
+  .claude/agents/qa-browser-tester.md: sha256:e57d5a6c920672206872cec0617e797b4059b13d87719a5364e08a95b5b8b9c6
+  .claude/agents/researcher.md: sha256:dfdfb57af650e20a1a173ba1a769d727fe4afd9e89ed3bd4b19cd4e290c3f79a
+  .claude/agents/reviewer.md: sha256:bdc276926bf77265af49698249ad677c7db7e7089b74ebc0e8fd72977d4383a2
+  .claude/agents/ruby-coder.md: sha256:5b1f7bcbd1e12b194268d24889420654941008d1c58bca33bd350bab95a251f6
+  .claude/agents/ruby-reviewer.md: sha256:b2b17e1726394e1891c25be639380b60549b337e9fd10dee77f0ae13f5b562de
+  .claude/agents/screenshot-guardian.md: sha256:edf9e8326306713be1095c0c3625af16b81cedb47bdf6c741e87e813a6dba723
+  .claude/agents/tester.md: sha256:7d59c5295ea98f66219f7aa7b4b861d4a3b104a3978f60cb36331426b34924ce
+  .claude/agents/tutorial-engineer.md: sha256:7dc7cbd1422abc2398e0982637ed770f4837048bd58a872ac8123257056b939b
+  .claude/agents/ui-problem-diagnosis.md: sha256:08910a672753738c73de4ea7b6635ea8966e2e8fecb0835c894f37c12f67903b
+  .claude/agents/ux-browser-validator.md: sha256:9d311911585106b65f15a7aae0dda2e82388b785d3e13608d356cb93fdea0ffe
+  .claude/commands/blog-pipeline.md: sha256:2d19b9a9fafcfd0ba1be79ec2ce2003c6e5230a8353e140516e4c74c05f8d53f
+  .claude/commands/cover-image.md: sha256:aa2235f48abfe6c6c61c43f8c133a0bee2176892622028b2ce1baaa04ab48ba4
+  .claude/commands/research.md: sha256:ebf8a17af5b5f66661fa827d534380c72327d3d57ebcd684e7690fe7e4afe73f
+  .claude/rules/css-consolidation.md: sha256:7664778a092d6aae023bf9630b444355074d225de6c2fa963c72a5102e5d523b
+  .claude/rules/icp-target.md: sha256:bc9afaac38691997e0632a78f7936473899ce16a2d959048a91e9ddd3fb5f5d0
+  .claude/rules/project-rules.md: sha256:26cdb09fdcf39ee9cc2317dd715f837bc21c6f64f38b2c19289ad000da0868c4
+  .claude/rules/tdd-enforcement.md: sha256:bf9eae071c503890569a36dada8efa53a9adeec7bf8ec95194908f6ba58707c6
+  .claude/rules/voice-guide.md: sha256:3dcf04d4b81bfd441795841e75c7d0b1b3dfd0c2e6ddebf8e8e680563c2954a9
+  .claude/skills/blog-post-pipeline/SKILL.md: sha256:f169ad41677c3e50d367f8edbd3b765d8f853dd967a70201e85e79ffe4b31356
+  .claude/skills/content-voice-enforcement/SKILL.md: sha256:0005acf787979ed0935833437e5225fc893a4c8ee5b206e19f81df0ebafb56f6
+  .claude/skills/cover-image-generation/SKILL.md: sha256:1b386045354b231a9575d682e9539e4244362ba686877f49985f1914002a1314
+  .claude/skills/css-bem-migration/SKILL.md: sha256:cfd51b99fefc745d40c5fa4cb7ea1755cea6b5d1f9ee4ea819c0cfd78f9fb85c
+  .claude/skills/visual-regression-testing/SKILL.md: sha256:952575d548ed20abce7e08f79bec4ba53a4c72987de0b507520069d332f8e10b
+  .codex/agents/analyst.toml: sha256:70faf83cc0b57308c3a9a3c79302a9096beee6d2db2f323777b84854d2dbbf05
+  .codex/agents/build-monitor.toml: sha256:e7021093abb12c1f3b1ac14be07f1addf8b681f42f6360c2c4cdb7f4180f5094
+  .codex/agents/coder.toml: sha256:a2f7e4444d0076895f57b43f1ab67d100ed00e426df2a89ddc5b60ce8fac7112
+  .codex/agents/content-creator.toml: sha256:a97e66aa5a901ebbef0065907a6bc100a512c52d21a9f6a16c561afe7f44b95c
+  .codex/agents/content-marketer.toml: sha256:2dd949cc86ee343993cbcca1988ee39ed1a7e58c452cfbab74928a7079c45848
+  .codex/agents/hugo-expert.toml: sha256:a2f3439af2851fbc01d5bbccc93b6977364fc44bb036b509b666373dd8240de1
+  .codex/agents/planner.toml: sha256:209afbc2422f29ab6d3fd2ebc2223ca6d527031c90d38a7f553387956a9026d4
+  .codex/agents/qa-browser-tester.toml: sha256:2b1dd26a13533bf447b32c78fa5c1a6a97e5c37b2bdbee148b0d16cdaeb65d11
+  .codex/agents/researcher.toml: sha256:bb69369bb32594d07c22f6a9e37988a6d6901843897cad9582837de8bce03865
+  .codex/agents/reviewer.toml: sha256:06759aee4c750db655fb32e6244d28ddc1ab0d43f788ba8fd5e50b01e21e618b
+  .codex/agents/ruby-coder.toml: sha256:a92df95c26135d9fcddf22b078ea542f5827b372d1b812c358e66030a4eda9f9
+  .codex/agents/ruby-reviewer.toml: sha256:36141dda08de3a4d95f005aefd099033b82757577a0c2bd1a9b3e95e5529f363
+  .codex/agents/screenshot-guardian.toml: sha256:07e01540221422d24a6cbce3bf6106c5785486007d5e5c5514ea676fc9c37da0
+  .codex/agents/seo-specialist.toml: sha256:f1200560cb238197e240a751f653f7c2f3c9b06e2f3937302d30be7bd8eb276a
+  .codex/agents/tester.toml: sha256:bf2306734665b4df060d36ac148cfa92e5f3511a8826cb24fe1da4e9d27daa1d
+  .codex/agents/tutorial-engineer.toml: sha256:8fb0a9eb7479ae1f61dff8479633c1fa5a0e6c6c21a3fc198be18e83ce465c7f
+  .codex/agents/ui-problem-diagnosis.toml: sha256:49cdc2fd50eced91a0699db6608e8071ce2620ddb65408eb2d6d3e970abde27c
+  .codex/agents/ux-browser-validator.toml: sha256:83564d3fa2c23ce43a421182e0fe175a9b1d91330619ccda57054fa83aeaaeb2
+  .cursor/agents/analyst.md: sha256:fd6b3b5c78aec9a2521e6e39f4256598e9aa0d01dedcbb2357a44cf5d4caf3e6
+  .cursor/agents/build-monitor.md: sha256:ec44db48151179313850daae620f1ef503b385466355a7c336fd21e8774efcbe
+  .cursor/agents/coder.md: sha256:60533575db22019d7a1a513e6efe854c5b59d1859f146d38132b123ffbb4aa17
+  .cursor/agents/content-creator.md: sha256:0fdda3cdc0dab7f4bc5c574a709c7e9374f6cc7b3ba1dcb072954390e33c8722
+  .cursor/agents/content-marketer.md: sha256:1cd7ee611816579e75a1c6c2ea8ce9a20a4712010da5b59848913a08b4d43ffe
+  .cursor/agents/hugo-expert.md: sha256:3460dd0cf80548d9a423cac015e2e3ecee2829a2e9d9773da5bbf465b1491162
+  .cursor/agents/planner.md: sha256:7dd9e396043bdd52989698f25e47c9cd1dc75b901677388ab062ff3237737706
+  .cursor/agents/qa-browser-tester.md: sha256:e57d5a6c920672206872cec0617e797b4059b13d87719a5364e08a95b5b8b9c6
+  .cursor/agents/researcher.md: sha256:dfdfb57af650e20a1a173ba1a769d727fe4afd9e89ed3bd4b19cd4e290c3f79a
+  .cursor/agents/reviewer.md: sha256:bdc276926bf77265af49698249ad677c7db7e7089b74ebc0e8fd72977d4383a2
+  .cursor/agents/ruby-coder.md: sha256:5b1f7bcbd1e12b194268d24889420654941008d1c58bca33bd350bab95a251f6
+  .cursor/agents/ruby-reviewer.md: sha256:b2b17e1726394e1891c25be639380b60549b337e9fd10dee77f0ae13f5b562de
+  .cursor/agents/screenshot-guardian.md: sha256:edf9e8326306713be1095c0c3625af16b81cedb47bdf6c741e87e813a6dba723
+  .cursor/agents/seo-specialist.md: sha256:2d9976d55381dae4fa25b0099eb5a2c1aa1ca863407a32e11a464c543bf7a649
+  .cursor/agents/tester.md: sha256:7d59c5295ea98f66219f7aa7b4b861d4a3b104a3978f60cb36331426b34924ce
+  .cursor/agents/tutorial-engineer.md: sha256:7dc7cbd1422abc2398e0982637ed770f4837048bd58a872ac8123257056b939b
+  .cursor/agents/ui-problem-diagnosis.md: sha256:08910a672753738c73de4ea7b6635ea8966e2e8fecb0835c894f37c12f67903b
+  .cursor/agents/ux-browser-validator.md: sha256:9d311911585106b65f15a7aae0dda2e82388b785d3e13608d356cb93fdea0ffe
+  .cursor/commands/blog-pipeline.md: sha256:2d19b9a9fafcfd0ba1be79ec2ce2003c6e5230a8353e140516e4c74c05f8d53f
+  .cursor/commands/cover-image.md: sha256:aa2235f48abfe6c6c61c43f8c133a0bee2176892622028b2ce1baaa04ab48ba4
+  .cursor/commands/research.md: sha256:ebf8a17af5b5f66661fa827d534380c72327d3d57ebcd684e7690fe7e4afe73f
+  .cursor/rules/css-consolidation.mdc: sha256:54c0515b4056c624929420108a99a854ae9aac7a5bcc3d8ce33bb2cc2254729b
+  .cursor/rules/icp-target.mdc: sha256:d374638f65f105e0432d862a194092c42351b2da040e9e22187203ebccc99dcc
+  .cursor/rules/project-rules.mdc: sha256:904d478370120c6001aa8b9d7d26c0d0854708aff7cdf073d10ab41ba0338ef9
+  .cursor/rules/tdd-enforcement.mdc: sha256:6fc26c5fbfb1e62962fdefae8e5d0436708b08a2e7c9d1e06e626a8e7da4d405
+  .cursor/rules/voice-guide.mdc: sha256:23df966ec8094c029016120806ca45ddcbc498eee843adb5647be0eacac51187
+  .gemini/commands/blog-pipeline.toml: sha256:9c222fd62bfeac3e28c2759198faf41c6e6bb092daa21625767a4ba22a33cf2f
+  .gemini/commands/cover-image.toml: sha256:e76ee8cafc142c51d38060f3b2d2d18dd782328d5cb9bff9166cffdbaf0d620e
+  .gemini/commands/research.toml: sha256:f552a11bd5825cca6ec95c3bfc211e2d3831e5386a60f2c94e8dd725d933f5a5
+  .github/agents/analyst.agent.md: sha256:fd6b3b5c78aec9a2521e6e39f4256598e9aa0d01dedcbb2357a44cf5d4caf3e6
+  .github/agents/build-monitor.agent.md: sha256:ec44db48151179313850daae620f1ef503b385466355a7c336fd21e8774efcbe
+  .github/agents/coder.agent.md: sha256:60533575db22019d7a1a513e6efe854c5b59d1859f146d38132b123ffbb4aa17
+  .github/agents/content-creator.agent.md: sha256:0fdda3cdc0dab7f4bc5c574a709c7e9374f6cc7b3ba1dcb072954390e33c8722
+  .github/agents/content-marketer.agent.md: sha256:1cd7ee611816579e75a1c6c2ea8ce9a20a4712010da5b59848913a08b4d43ffe
+  .github/agents/hugo-expert.agent.md: sha256:3460dd0cf80548d9a423cac015e2e3ecee2829a2e9d9773da5bbf465b1491162
+  .github/agents/planner.agent.md: sha256:7dd9e396043bdd52989698f25e47c9cd1dc75b901677388ab062ff3237737706
+  .github/agents/qa-browser-tester.agent.md: sha256:e57d5a6c920672206872cec0617e797b4059b13d87719a5364e08a95b5b8b9c6
+  .github/agents/researcher.agent.md: sha256:dfdfb57af650e20a1a173ba1a769d727fe4afd9e89ed3bd4b19cd4e290c3f79a
+  .github/agents/reviewer.agent.md: sha256:bdc276926bf77265af49698249ad677c7db7e7089b74ebc0e8fd72977d4383a2
+  .github/agents/ruby-coder.agent.md: sha256:5b1f7bcbd1e12b194268d24889420654941008d1c58bca33bd350bab95a251f6
+  .github/agents/ruby-reviewer.agent.md: sha256:b2b17e1726394e1891c25be639380b60549b337e9fd10dee77f0ae13f5b562de
+  .github/agents/screenshot-guardian.agent.md: sha256:edf9e8326306713be1095c0c3625af16b81cedb47bdf6c741e87e813a6dba723
+  .github/agents/seo-specialist.agent.md: sha256:2d9976d55381dae4fa25b0099eb5a2c1aa1ca863407a32e11a464c543bf7a649
+  .github/agents/tester.agent.md: sha256:7d59c5295ea98f66219f7aa7b4b861d4a3b104a3978f60cb36331426b34924ce
+  .github/agents/tutorial-engineer.agent.md: sha256:7dc7cbd1422abc2398e0982637ed770f4837048bd58a872ac8123257056b939b
+  .github/agents/ui-problem-diagnosis.agent.md: sha256:08910a672753738c73de4ea7b6635ea8966e2e8fecb0835c894f37c12f67903b
+  .github/agents/ux-browser-validator.agent.md: sha256:9d311911585106b65f15a7aae0dda2e82388b785d3e13608d356cb93fdea0ffe
+  .github/instructions/css-consolidation.instructions.md: sha256:7664778a092d6aae023bf9630b444355074d225de6c2fa963c72a5102e5d523b
+  .github/instructions/icp-target.instructions.md: sha256:bc9afaac38691997e0632a78f7936473899ce16a2d959048a91e9ddd3fb5f5d0
+  .github/instructions/project-rules.instructions.md: sha256:26cdb09fdcf39ee9cc2317dd715f837bc21c6f64f38b2c19289ad000da0868c4
+  .github/instructions/tdd-enforcement.instructions.md: sha256:bf9eae071c503890569a36dada8efa53a9adeec7bf8ec95194908f6ba58707c6
+  .github/instructions/voice-guide.instructions.md: sha256:3dcf04d4b81bfd441795841e75c7d0b1b3dfd0c2e6ddebf8e8e680563c2954a9
+  .github/prompts/blog-pipeline.prompt.md: sha256:f960e4824d90ebf06b5395bbf5a964c47a5edc3b207e0f43a78d5f4dfd38ec5e
+  .github/prompts/cover-image.prompt.md: sha256:20ccbc871c4028307cca49b1d8e7521ada2c9b7489be56fbeb66aabdd66ad857
+  .github/prompts/research.prompt.md: sha256:19878f66c1a89e38cba9c9078248d9ec08fda0001eaaae992d0a42ca81cb62e9
+  .kiro/skills/blog-post-pipeline/SKILL.md: sha256:f169ad41677c3e50d367f8edbd3b765d8f853dd967a70201e85e79ffe4b31356
+  .kiro/skills/content-voice-enforcement/SKILL.md: sha256:0005acf787979ed0935833437e5225fc893a4c8ee5b206e19f81df0ebafb56f6
+  .kiro/skills/cover-image-generation/SKILL.md: sha256:1b386045354b231a9575d682e9539e4244362ba686877f49985f1914002a1314
+  .kiro/skills/css-bem-migration/SKILL.md: sha256:cfd51b99fefc745d40c5fa4cb7ea1755cea6b5d1f9ee4ea819c0cfd78f9fb85c
+  .kiro/skills/visual-regression-testing/SKILL.md: sha256:952575d548ed20abce7e08f79bec4ba53a4c72987de0b507520069d332f8e10b
+  .kiro/steering/css-consolidation.md: sha256:8f622fae6291d1ce4c906a170a1f0a5f9cc03263ea08130aaf4e414662345388
+  .kiro/steering/icp-target.md: sha256:772ef808ed9112588aabf3d51c2ff1bcc2b19555491a611d60ce0d3981e4fd44
+  .kiro/steering/project-rules.md: sha256:df049b1e0e4f976ad1ca11586164c237f90dfa79c4178fd1f00ddf3cabd1ebf2
+  .kiro/steering/tdd-enforcement.md: sha256:1c6267029066dbaa4a96f2c7a8cdcff976806238aef5c8ad281446a4d2b3f4b2
+  .kiro/steering/voice-guide.md: sha256:f17939a682fc28734fa3992e8f10576c73c3074eebf7a094d615ae0954b5c335
+  .opencode/agents/analyst.md: sha256:fd6b3b5c78aec9a2521e6e39f4256598e9aa0d01dedcbb2357a44cf5d4caf3e6
+  .opencode/agents/build-monitor.md: sha256:ec44db48151179313850daae620f1ef503b385466355a7c336fd21e8774efcbe
+  .opencode/agents/coder.md: sha256:60533575db22019d7a1a513e6efe854c5b59d1859f146d38132b123ffbb4aa17
+  .opencode/agents/content-creator.md: sha256:0fdda3cdc0dab7f4bc5c574a709c7e9374f6cc7b3ba1dcb072954390e33c8722
+  .opencode/agents/content-marketer.md: sha256:1cd7ee611816579e75a1c6c2ea8ce9a20a4712010da5b59848913a08b4d43ffe
+  .opencode/agents/hugo-expert.md: sha256:3460dd0cf80548d9a423cac015e2e3ecee2829a2e9d9773da5bbf465b1491162
+  .opencode/agents/planner.md: sha256:7dd9e396043bdd52989698f25e47c9cd1dc75b901677388ab062ff3237737706
+  .opencode/agents/qa-browser-tester.md: sha256:e57d5a6c920672206872cec0617e797b4059b13d87719a5364e08a95b5b8b9c6
+  .opencode/agents/researcher.md: sha256:dfdfb57af650e20a1a173ba1a769d727fe4afd9e89ed3bd4b19cd4e290c3f79a
+  .opencode/agents/reviewer.md: sha256:bdc276926bf77265af49698249ad677c7db7e7089b74ebc0e8fd72977d4383a2
+  .opencode/agents/ruby-coder.md: sha256:5b1f7bcbd1e12b194268d24889420654941008d1c58bca33bd350bab95a251f6
+  .opencode/agents/ruby-reviewer.md: sha256:b2b17e1726394e1891c25be639380b60549b337e9fd10dee77f0ae13f5b562de
+  .opencode/agents/screenshot-guardian.md: sha256:edf9e8326306713be1095c0c3625af16b81cedb47bdf6c741e87e813a6dba723
+  .opencode/agents/seo-specialist.md: sha256:2d9976d55381dae4fa25b0099eb5a2c1aa1ca863407a32e11a464c543bf7a649
+  .opencode/agents/tester.md: sha256:7d59c5295ea98f66219f7aa7b4b861d4a3b104a3978f60cb36331426b34924ce
+  .opencode/agents/tutorial-engineer.md: sha256:7dc7cbd1422abc2398e0982637ed770f4837048bd58a872ac8123257056b939b
+  .opencode/agents/ui-problem-diagnosis.md: sha256:08910a672753738c73de4ea7b6635ea8966e2e8fecb0835c894f37c12f67903b
+  .opencode/agents/ux-browser-validator.md: sha256:9d311911585106b65f15a7aae0dda2e82388b785d3e13608d356cb93fdea0ffe
+  .opencode/commands/blog-pipeline.md: sha256:2d19b9a9fafcfd0ba1be79ec2ce2003c6e5230a8353e140516e4c74c05f8d53f
+  .opencode/commands/cover-image.md: sha256:aa2235f48abfe6c6c61c43f8c133a0bee2176892622028b2ce1baaa04ab48ba4
+  .opencode/commands/research.md: sha256:ebf8a17af5b5f66661fa827d534380c72327d3d57ebcd684e7690fe7e4afe73f
+  .windsurf/rules/css-consolidation.md: sha256:dd6f393a0c5e0f5d4a2f701c598b108d76a5210b91fba389cac6a9ca3c73ccc6
+  .windsurf/rules/icp-target.md: sha256:0162f023b9084fb86430df3754024c5a549112d81bd33b8b3191f8722d75233f
+  .windsurf/rules/project-rules.md: sha256:7a9408331ce330fe113d9b388b1f55efc3ef6ecdb270637015dd7ce9ef75ded7
+  .windsurf/rules/tdd-enforcement.md: sha256:8143c2bb851b5ead6008e3a5654b50f51df10299bd57ab21911aa95317cce1e8
+  .windsurf/rules/voice-guide.md: sha256:532c3202b8bfc9227e5de97b35ba0bac6f6b44fa4b9d12737e9c44754c9b90f4
+  .windsurf/workflows/blog-pipeline.md: sha256:2d19b9a9fafcfd0ba1be79ec2ce2003c6e5230a8353e140516e4c74c05f8d53f
+  .windsurf/workflows/cover-image.md: sha256:aa2235f48abfe6c6c61c43f8c133a0bee2176892622028b2ce1baaa04ab48ba4
+  .windsurf/workflows/research.md: sha256:ebf8a17af5b5f66661fa827d534380c72327d3d57ebcd684e7690fe7e4afe73f
diff --git a/apm.yml b/apm.yml
new file mode 100644
index 000000000..a3014a8e3
--- /dev/null
+++ b/apm.yml
@@ -0,0 +1,135 @@
+# =============================================================================
+# APM Manifest — JetThoughts Site
+# =============================================================================
+# This repo is a PRODUCER: it defines canonical agents, skills, instructions,
+# and prompts reusable across all JetThoughts repos.
+#
+# Tools auto-configured by APM v0.24.0:
+#   Claude Code, GitHub Copilot, Codex CLI, Cursor, Gemini CLI, Windsurf,
+#   Kiro, OpenCode, Antigravity (AGENTS.md+skills).
+#
+# Tools configured manually (APM doesn't support these targets yet):
+#   Devin, Codebuff, Qwen Code, Junie — use thin-wrapper files at root.
+#
+# Quick start:
+#   apm install       # Compile .apm/ → tool configs
+#   make mcp-up       # Start persistent MCP backends (ollama, searxng, milvus)
+#   code .            # Open in any AI tool — same config everywhere
+# =============================================================================
+
+name: jetthoughts/site
+version: 0.1.0
+description: >
+  JetThoughts website — Hugo static site blog with AI-assisted content
+  pipeline, BEM CSS architecture, visual regression testing, and TDD
+  enforcement. Exports reusable voice guides, blog-pipeline skills,
+  CSS-consolidation patterns, and agent definitions.
+
+# ---------------------------------------------------------------------------
+# DEPENDENCIES — what this project consumes
+# ---------------------------------------------------------------------------
+dependencies:
+
+  # --- APM Packages ---
+  apm: []
+
+  # --- MCP Servers ---
+  mcp:
+
+    # Stdio-based (auto-started by AI tool's MCP client)
+    # claude-flow-memory intentionally excluded — claude-flow plugin dependency
+
+    - name: claude-context
+      registry: false
+      transport: stdio
+      command: npx
+      args:
+        - "-y"
+        - "@anthropic/claude-context"
+
+    - name: context7
+      registry: false
+      transport: stdio
+      command: npx
+      args:
+        - "-y"
+        - "@upstash/context7-mcp"
+
+    - name: brave-search
+      registry: false
+      transport: stdio
+      command: npx
+      args:
+        - "-y"
+        - "@anthropic/brave-search-mcp"
+      env:
+        BRAVE_API_KEY: "${BRAVE_API_KEY}"
+
+    - name: lean-ctx
+      registry: false
+      transport: stdio
+      command: npx
+      args:
+        - "-y"
+        - "lean-ctx-mcp"
+
+    - name: chrome-devtools
+      registry: false
+      transport: stdio
+      command: npx
+      args:
+        - "-y"
+        - "@anthropic/chrome-devtools-mcp"
+
+    - name: package-search
+      registry: false
+      transport: stdio
+      command: npx
+      args:
+        - "-y"
+        - "package-search-mcp"
+
+    - name: github-mcp-server
+      registry: false
+      transport: stdio
+      command: npx
+      args:
+        - "-y"
+        - "@anthropic/github-mcp-server"
+      env:
+        GITHUB_TOKEN: "${GITHUB_TOKEN}"
+
+    # HTTP-based persistent backends (started via `make mcp-up`)
+    # These run as Docker services defined in compose.mcp.yml
+
+    - name: searxng
+      registry: false
+      transport: http
+      url: http://localhost:8080
+
+    - name: ollama
+      registry: false
+      transport: http
+      url: http://localhost:11434
+
+    - name: milvus
+      registry: false
+      transport: http
+      url: http://localhost:19530
+
+# ---------------------------------------------------------------------------
+# TOOL TARGETS — which tools APM compiles configuration for
+# ---------------------------------------------------------------------------
+# All share .apm/instructions/project-rules.instructions.md.
+# Thin-wrapper files (AGENTS.md, GEMINI.md, QWEN.md, .junie/AGENTS.md)
+# serve as fallback for tools without APM target support.
+
+targets:
+  - claude
+  - copilot
+  - codex
+  - cursor
+  - gemini
+  - opencode
+  - windsurf
+  - kiro
diff --git a/compose.mcp.yml b/compose.mcp.yml
new file mode 100644
index 000000000..050b3a368
--- /dev/null
+++ b/compose.mcp.yml
@@ -0,0 +1,67 @@
+# compose.mcp.yml — MCP Infrastructure for JetThoughts Site
+#
+# Persistent backends for MCP servers declared in apm.yml → dependencies.mcp.
+# Started via: make mcp-up    (docker compose -f compose.mcp.yml up -d)
+# Stopped via: make mcp-down  (docker compose -f compose.mcp.yml down)
+#
+# These run per-repo — clone the repo, run make mcp-up, done.
+# Stdio-based MCPs (claude-context, context7, brave-search, etc.) are
+# auto-started by the AI tool's MCP client — no compose entry needed.
+# ==========================================================================
+
+services:
+  # ---- SearXNG — Privacy-respecting meta-search engine ----
+  # Used by: searxng MCP server → web research for blog posts
+  # Endpoint: http://localhost:8080
+  searxng:
+    image: searxng/searxng:latest
+    ports:
+      - "8080:8080"
+    volumes:
+      - searxng_data:/etc/searxng
+    environment:
+      - SEARXNG_BASE_URL=http://localhost:8080/
+    restart: unless-stopped
+
+  # ---- Milvus — Vector database for semantic search ----
+  # Used by: claude-context MCP server (storage backend)
+  # Endpoint: http://localhost:19530
+  milvus:
+    image: milvusdb/milvus:v2.4.0
+    ports:
+      - "19530:19530"
+      - "9091:9091"
+    environment:
+      ETCD_ENDPOINTS: etcd:2379
+      MINIO_ADDRESS: minio:9000
+    volumes:
+      - milvus_data:/var/lib/milvus
+    restart: unless-stopped
+    depends_on:
+      - etcd
+      - minio
+
+  etcd:
+    image: quay.io/coreos/etcd:v3.5.5
+    environment:
+      ETCD_AUTO_COMPACTION_MODE: revision
+      ETCD_AUTO_COMPACTION_RETENTION: "1000"
+      ETCD_QUOTA_BACKEND_BYTES: "4294967296"
+    volumes:
+      - etcd_data:/etcd
+    command: etcd -advertise-client-urls=http://127.0.0.1:2379 -listen-client-urls http://0.0.0.0:2379
+
+  minio:
+    image: minio/minio:RELEASE.2023-03-20T20-16-18Z
+    environment:
+      MINIO_ACCESS_KEY: minioadmin
+      MINIO_SECRET_KEY: minioadmin
+    volumes:
+      - minio_data:/minio_data
+    command: minio server /minio_data --console-address ":9001"
+
+volumes:
+  searxng_data:
+  milvus_data:
+  etcd_data:
+  minio_data:
diff --git a/docs/mcp-user-setup.md b/docs/mcp-user-setup.md
new file mode 100644
index 000000000..68e738a67
--- /dev/null
+++ b/docs/mcp-user-setup.md
@@ -0,0 +1,40 @@
+# Per-Repo MCP Setup
+
+Persistent MCP backends (searxng, milvus) run as Docker services defined
+in `compose.mcp.yml`. One `make mcp-up` starts everything.
+
+## Prerequisites
+
+- Docker (Desktop or colima)
+- Homebrew (for ollama)
+
+## One-Time Setup
+
+```bash
+# Install system dependencies
+brew bundle                    # installs ollama, bun, hugo, rbenv, ruby-build
+
+# Pull Docker images (optional — compose will pull on first up)
+docker compose -f compose.mcp.yml pull
+```
+
+## Daily Use
+
+```bash
+make mcp-up      # Start ollama (brew) + searxng/milvus (docker compose)
+make mcp-status  # Health check
+make mcp-down    # Stop Docker services
+```
+
+## Architecture
+
+```
+apm.yml                          # Declares all MCP servers
+compose.mcp.yml                  # Docker services: searxng, milvus, etcd, minio
+Makefile (mcp-up/mcp-down)       # Orchestration: brew + docker compose
+Brewfile                         # System deps: ollama, bun, hugo, rbenv
+```
+
+- **Stdio MCPs** (claude-context, context7, brave-search, etc.): auto-started by the AI tool's MCP client. No infrastructure needed.
+- **Ollama**: managed via `brew services`. Started by `brew bundle` (restart_service: true).
+- **SearXNG + Milvus**: Docker services in `compose.mcp.yml`. Started by `make mcp-up`.
diff --git a/opencode.json b/opencode.json
new file mode 100644
index 000000000..333cc299d
--- /dev/null
+++ b/opencode.json
@@ -0,0 +1,88 @@
+{
+  "mcp": {
+    "claude-context": {
+      "type": "local",
+      "enabled": true,
+      "command": [
+        "npx",
+        "-y",
+        "@anthropic/claude-context"
+      ]
+    },
+    "context7": {
+      "type": "local",
+      "enabled": true,
+      "command": [
+        "npx",
+        "-y",
+        "@upstash/context7-mcp"
+      ]
+    },
+    "brave-search": {
+      "type": "local",
+      "enabled": true,
+      "command": [
+        "npx",
+        "-y",
+        "@anthropic/brave-search-mcp"
+      ],
+      "environment": {
+        "BRAVE_API_KEY": "${BRAVE_API_KEY}"
+      }
+    },
+    "lean-ctx": {
+      "type": "local",
+      "enabled": true,
+      "command": [
+        "npx",
+        "-y",
+        "lean-ctx-mcp"
+      ]
+    },
+    "chrome-devtools": {
+      "type": "local",
+      "enabled": true,
+      "command": [
+        "npx",
+        "-y",
+        "@anthropic/chrome-devtools-mcp"
+      ]
+    },
+    "package-search": {
+      "type": "local",
+      "enabled": true,
+      "command": [
+        "npx",
+        "-y",
+        "package-search-mcp"
+      ]
+    },
+    "github-mcp-server": {
+      "type": "local",
+      "enabled": true,
+      "command": [
+        "npx",
+        "-y",
+        "@anthropic/github-mcp-server"
+      ],
+      "environment": {
+        "GITHUB_TOKEN": "${GITHUB_TOKEN}"
+      }
+    },
+    "searxng": {
+      "type": "remote",
+      "enabled": true,
+      "url": "http://localhost:8080"
+    },
+    "ollama": {
+      "type": "remote",
+      "enabled": true,
+      "url": "http://localhost:11434"
+    },
+    "milvus": {
+      "type": "remote",
+      "enabled": true,
+      "url": "http://localhost:19530"
+    }
+  }
+}
\ No newline at end of file

From f4cd730fcf3bfe052c08e533098b623ff89427fc Mon Sep 17 00:00:00 2001
From: Paul Keen <pftg@Pauls-MacBook-Pro.local>
Date: Tue, 7 Jul 2026 23:30:29 -0400
Subject: [PATCH 2/6] fix(mcp): add milvus standalone command and fix health
 checks for colima
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

- Add command: ["milvus", "run", "standalone"] to compose.mcp.yml
  (milvus v2.4.0 requires explicit standalone entrypoint)
- Switch docker compose → docker-compose in Makefile targets
  (colima with standalone docker-cli lacks compose plugin)
- Fix milvus health check: use nc port check instead of curl /health
  (milvus v2.4.0 has no /health HTTP endpoint on :19530)
---
 Makefile        | 10 +++++-----
 compose.mcp.yml |  1 +
 2 files changed, 6 insertions(+), 5 deletions(-)

diff --git a/Makefile b/Makefile
index 2713bb30c..f2e4e1fb7 100644
--- a/Makefile
+++ b/Makefile
@@ -1,12 +1,12 @@
 # Makefile — JetThoughts Site
 #
 # APM declares MCP dependencies in apm.yml (what agents need).
-# This Makefile RUNS the persistent backends via docker compose + brew.
+# This Makefile RUNS the persistent backends via docker-compose + brew.
 # Stdio-based MCPs are auto-started by the AI tool — no target needed.
 #
 # Quick start:
 #   apm install      # Install agent primitives (instructions, agents, prompts)
-#   make mcp-up      # Start persistent MCP backends
+#   make mcp-up      # Start persistent MCP backends (via docker-compose)
 #   make mcp-status  # Verify everything is running
 #   code .           # Open in Claude/Copilot/Cursor/etc.
 
@@ -30,13 +30,13 @@ mcp-up:
 	@brew services start ollama 2>/dev/null && echo "   ✓ ollama started" || echo "   ! ollama not installed — run: brew install ollama"
 	@echo ""
 	# Docker-based MCP services (searxng, milvus, etcd, minio)
-	@docker compose -f compose.mcp.yml up -d
+	@docker-compose -f compose.mcp.yml up -d
 	@echo ""
 	@$(MAKE) mcp-status
 
 mcp-down:
 	@echo "==> Stopping MCP backends..."
-	@docker compose -f compose.mcp.yml down
+	@docker-compose -f compose.mcp.yml down
 	@echo "   ollama: left running (brew services stop ollama to stop)"
 	@echo "==> MCP backends stopped"
 
@@ -48,5 +48,5 @@ mcp-status:
 	@printf "   %-14s " "searxng :8080"; \
 		curl -s -o /dev/null -w "%{http_code}" http://localhost:8080/healthz 2>/dev/null | grep -q 200 && echo "✓ running" || echo "✗ down"
 	@printf "   %-14s " "milvus :19530"; \
-		curl -s -o /dev/null -w "%{http_code}" http://localhost:19530/health 2>/dev/null | grep -q 200 && echo "✓ running" || echo "✗ down"
+		nc -z -w 2 localhost 19530 2>/dev/null && echo "✓ running" || echo "✗ down"
 	@echo ""
diff --git a/compose.mcp.yml b/compose.mcp.yml
index 050b3a368..ab770642c 100644
--- a/compose.mcp.yml
+++ b/compose.mcp.yml
@@ -28,6 +28,7 @@ services:
   # Endpoint: http://localhost:19530
   milvus:
     image: milvusdb/milvus:v2.4.0
+    command: ["milvus", "run", "standalone"]
     ports:
       - "19530:19530"
       - "9091:9091"

From 33fb7682f691294f3cb9cf56b413f7ac2aa6e64e Mon Sep 17 00:00:00 2001
From: Paul Keen <pftg@Pauls-MacBook-Pro.local>
Date: Tue, 7 Jul 2026 23:37:29 -0400
Subject: [PATCH 3/6] chore(mcp): bump milvus to v2.6.19, etcd to v3.5.25,
 minio to 2024-05-28
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

- milvusdb/milvus: v2.4.0 → v2.6.19 (latest stable as of Jul 2026)
- quay.io/coreos/etcd: v3.5.5 → v3.5.25 (v2.6 compatible metadata schema)
- minio/minio: 2023-03-20 → 2024-05-28 (v2.6 compatible storage)

All versions match official milvus v2.6 standalone compose template.
Verified: containers start, healthz OK, searxng API returns results.
---
 compose.mcp.yml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/compose.mcp.yml b/compose.mcp.yml
index ab770642c..4675f2210 100644
--- a/compose.mcp.yml
+++ b/compose.mcp.yml
@@ -27,7 +27,7 @@ services:
   # Used by: claude-context MCP server (storage backend)
   # Endpoint: http://localhost:19530
   milvus:
-    image: milvusdb/milvus:v2.4.0
+    image: milvusdb/milvus:v2.6.19
     command: ["milvus", "run", "standalone"]
     ports:
       - "19530:19530"
@@ -43,7 +43,7 @@ services:
       - minio
 
   etcd:
-    image: quay.io/coreos/etcd:v3.5.5
+    image: quay.io/coreos/etcd:v3.5.25
     environment:
       ETCD_AUTO_COMPACTION_MODE: revision
       ETCD_AUTO_COMPACTION_RETENTION: "1000"
@@ -53,7 +53,7 @@ services:
     command: etcd -advertise-client-urls=http://127.0.0.1:2379 -listen-client-urls http://0.0.0.0:2379
 
   minio:
-    image: minio/minio:RELEASE.2023-03-20T20-16-18Z
+    image: minio/minio:RELEASE.2024-05-28T17-19-04Z
     environment:
       MINIO_ACCESS_KEY: minioadmin
       MINIO_SECRET_KEY: minioadmin

From a7c639e0644ce9f4cdc5ed6cb39a8683c38c39ee Mon Sep 17 00:00:00 2001
From: Paul Keen <pftg@Pauls-MacBook-Pro.local>
Date: Tue, 7 Jul 2026 23:42:21 -0400
Subject: [PATCH 4/6] feat(mcp): add healthchecks to etcd/minio with
 service_healthy dependency

- Add etcd healthcheck (etcdctl endpoint health) with start_period
- Add minio healthcheck (bash /dev/tcp port 9000 check) with start_period
  (minio image has no curl/wget/mc/pgrep, uses bash built-in TCP)
- Update milvus depends_on to condition: service_healthy
  Ensures etcd+minio are ready before milvus starts, eliminating
  startup race conditions that caused milvus crash loops.
---
 compose.mcp.yml | 18 ++++++++++++++++--
 1 file changed, 16 insertions(+), 2 deletions(-)

diff --git a/compose.mcp.yml b/compose.mcp.yml
index 4675f2210..0d16bbe92 100644
--- a/compose.mcp.yml
+++ b/compose.mcp.yml
@@ -39,8 +39,10 @@ services:
       - milvus_data:/var/lib/milvus
     restart: unless-stopped
     depends_on:
-      - etcd
-      - minio
+      etcd:
+        condition: service_healthy
+      minio:
+        condition: service_healthy
 
   etcd:
     image: quay.io/coreos/etcd:v3.5.25
@@ -51,6 +53,12 @@ services:
     volumes:
       - etcd_data:/etcd
     command: etcd -advertise-client-urls=http://127.0.0.1:2379 -listen-client-urls http://0.0.0.0:2379
+    healthcheck:
+      test: ["CMD", "etcdctl", "endpoint", "health"]
+      interval: 10s
+      timeout: 5s
+      retries: 5
+      start_period: 10s
 
   minio:
     image: minio/minio:RELEASE.2024-05-28T17-19-04Z
@@ -60,6 +68,12 @@ services:
     volumes:
       - minio_data:/minio_data
     command: minio server /minio_data --console-address ":9001"
+    healthcheck:
+      test: ["CMD", "bash", "-c", "echo > /dev/tcp/localhost/9000"]
+      interval: 10s
+      timeout: 5s
+      retries: 5
+      start_period: 10s
 
 volumes:
   searxng_data:

From 05fdc58f0dc67bf305089443268bd43844df1e97 Mon Sep 17 00:00:00 2001
From: Paul Keen <pftg@Pauls-MacBook-Pro.local>
Date: Tue, 7 Jul 2026 23:45:08 -0400
Subject: [PATCH 5/6] feat(mcp): add seed searxng settings.yml with JSON API
 enabled

- Create .dev/searxng/settings.yml enabling html+json search formats
- Bind-mount read-only into searxng container in compose.mcp.yml
- Fixes: JSON API was previously ephemeral (written to Docker volume),
  lost on docker-compose down -v or volume prune
- JSON at :8080/search?q=X&format=json now works out of the box
  on fresh clone after make mcp-up
---
 .dev/searxng/settings.yml | 13 +++++++++++++
 compose.mcp.yml           |  1 +
 2 files changed, 14 insertions(+)
 create mode 100644 .dev/searxng/settings.yml

diff --git a/.dev/searxng/settings.yml b/.dev/searxng/settings.yml
new file mode 100644
index 000000000..fe9905b64
--- /dev/null
+++ b/.dev/searxng/settings.yml
@@ -0,0 +1,13 @@
+# SearXNG settings — JetThoughts local MCP instance
+# Mounted read-only via compose.mcp.yml
+use_default_settings: true
+
+search:
+  formats:
+    - html
+    - json
+
+server:
+  secret_key: "1oOHoKQ97m1uS9AjDH3fDVr0ZeQCyJjc"
+  bind_address: "0.0.0.0"
+  image_proxy: true
diff --git a/compose.mcp.yml b/compose.mcp.yml
index 0d16bbe92..b2dbfbbe0 100644
--- a/compose.mcp.yml
+++ b/compose.mcp.yml
@@ -19,6 +19,7 @@ services:
       - "8080:8080"
     volumes:
       - searxng_data:/etc/searxng
+      - ./.dev/searxng/settings.yml:/etc/searxng/settings.yml:ro
     environment:
       - SEARXNG_BASE_URL=http://localhost:8080/
     restart: unless-stopped

From 19484183659bf958e0dddf65e236a6c81b69e364 Mon Sep 17 00:00:00 2001
From: Paul Keen <pftg@Pauls-MacBook-Pro.local>
Date: Tue, 7 Jul 2026 23:52:57 -0400
Subject: [PATCH 6/6] feat(mcp): switch claude-context to
 @zilliz/claude-context-mcp with local Milvus+Ollama

- Migrate from @anthropic/claude-context (GOB file storage) to
  @zilliz/claude-context-mcp@latest (Milvus vector backend)
- Configure Ollama embeddings (nomic-embed-text, dim 768)
- Point to local Milvus (localhost:19530, root:Milvus)
- Update all 8 MCP config files + docs/mcp-user-setup.md

Prerequisites verified:
- Package @zilliz/claude-context-mcp exists (v0.1.15)
- Ollama has nomic-embed-text model
- Milvus (:9091/healthz) and Ollama (:11434/api/tags) reachable
- Server starts without errors
---
 .codex/config.toml      |  8 +++-
 .cursor/mcp.json        | 12 +++++-
 .gemini/settings.json   | 12 +++++-
 .kiro/settings/mcp.json | 12 +++++-
 .mcp.json               | 86 +++++++++++++++++++++++++++++++++++++++++
 apm.lock.yaml           |  9 ++++-
 apm.yml                 |  9 ++++-
 docs/mcp-user-setup.md  |  3 +-
 opencode.json           | 12 +++++-
 9 files changed, 151 insertions(+), 12 deletions(-)
 create mode 100644 .mcp.json

diff --git a/.codex/config.toml b/.codex/config.toml
index be084c500..e7b30822e 100644
--- a/.codex/config.toml
+++ b/.codex/config.toml
@@ -1,6 +1,6 @@
 [mcp_servers.claude-context]
 command = "npx"
-args = [ "-y", "@anthropic/claude-context",]
+args = [ "-y", "@zilliz/claude-context-mcp@latest",]
 id = ""
 
 [mcp_servers.context7]
@@ -34,6 +34,12 @@ args = [ "-y", "@anthropic/github-mcp-server",]
 id = ""
 
 [mcp_servers.claude-context.env]
+EMBEDDING_PROVIDER = "Ollama"
+EMBEDDING_MODEL = "nomic-embed-text"
+OLLAMA_HOST = "http://localhost:11434"
+EMBEDDING_DIMENSION = "768"
+MILVUS_ADDRESS = "http://localhost:19530"
+MILVUS_TOKEN = "root:Milvus"
 
 [mcp_servers.context7.env]
 
diff --git a/.cursor/mcp.json b/.cursor/mcp.json
index 7d37ea206..d2cf02837 100644
--- a/.cursor/mcp.json
+++ b/.cursor/mcp.json
@@ -5,8 +5,16 @@
       "command": "npx",
       "args": [
         "-y",
-        "@anthropic/claude-context"
-      ]
+        "@zilliz/claude-context-mcp@latest"
+      ],
+      "env": {
+        "EMBEDDING_PROVIDER": "Ollama",
+        "EMBEDDING_MODEL": "nomic-embed-text",
+        "OLLAMA_HOST": "http://localhost:11434",
+        "EMBEDDING_DIMENSION": "768",
+        "MILVUS_ADDRESS": "http://localhost:19530",
+        "MILVUS_TOKEN": "root:Milvus"
+      }
     },
     "context7": {
       "type": "stdio",
diff --git a/.gemini/settings.json b/.gemini/settings.json
index 46037b8c2..120a7833b 100644
--- a/.gemini/settings.json
+++ b/.gemini/settings.json
@@ -4,8 +4,16 @@
       "command": "npx",
       "args": [
         "-y",
-        "@anthropic/claude-context"
-      ]
+        "@zilliz/claude-context-mcp@latest"
+      ],
+      "env": {
+        "EMBEDDING_PROVIDER": "Ollama",
+        "EMBEDDING_MODEL": "nomic-embed-text",
+        "OLLAMA_HOST": "http://localhost:11434",
+        "EMBEDDING_DIMENSION": "768",
+        "MILVUS_ADDRESS": "http://localhost:19530",
+        "MILVUS_TOKEN": "root:Milvus"
+      }
     },
     "context7": {
       "command": "npx",
diff --git a/.kiro/settings/mcp.json b/.kiro/settings/mcp.json
index 368800ab2..706f0d207 100644
--- a/.kiro/settings/mcp.json
+++ b/.kiro/settings/mcp.json
@@ -4,8 +4,16 @@
       "command": "npx",
       "args": [
         "-y",
-        "@anthropic/claude-context"
-      ]
+        "@zilliz/claude-context-mcp@latest"
+      ],
+      "env": {
+        "EMBEDDING_PROVIDER": "Ollama",
+        "EMBEDDING_MODEL": "nomic-embed-text",
+        "OLLAMA_HOST": "http://localhost:11434",
+        "EMBEDDING_DIMENSION": "768",
+        "MILVUS_ADDRESS": "http://localhost:19530",
+        "MILVUS_TOKEN": "root:Milvus"
+      }
     },
     "context7": {
       "command": "npx",
diff --git a/.mcp.json b/.mcp.json
new file mode 100644
index 000000000..3bbac4feb
--- /dev/null
+++ b/.mcp.json
@@ -0,0 +1,86 @@
+{
+  "mcpServers": {
+    "claude-context": {
+      "type": "stdio",
+      "command": "npx",
+      "args": [
+        "-y",
+        "@zilliz/claude-context-mcp@latest"
+      ],
+      "env": {
+        "EMBEDDING_PROVIDER": "Ollama",
+        "EMBEDDING_MODEL": "nomic-embed-text",
+        "OLLAMA_HOST": "http://localhost:11434",
+        "EMBEDDING_DIMENSION": "768",
+        "MILVUS_ADDRESS": "http://localhost:19530",
+        "MILVUS_TOKEN": "root:Milvus"
+      }
+    },
+    "context7": {
+      "type": "stdio",
+      "command": "npx",
+      "args": [
+        "-y",
+        "@upstash/context7-mcp"
+      ]
+    },
+    "brave-search": {
+      "type": "stdio",
+      "command": "npx",
+      "env": {
+        "BRAVE_API_KEY": "${BRAVE_API_KEY}"
+      },
+      "args": [
+        "-y",
+        "@anthropic/brave-search-mcp"
+      ]
+    },
+    "lean-ctx": {
+      "type": "stdio",
+      "command": "npx",
+      "args": [
+        "-y",
+        "lean-ctx-mcp"
+      ]
+    },
+    "chrome-devtools": {
+      "type": "stdio",
+      "command": "npx",
+      "args": [
+        "-y",
+        "@anthropic/chrome-devtools-mcp"
+      ]
+    },
+    "package-search": {
+      "type": "stdio",
+      "command": "npx",
+      "args": [
+        "-y",
+        "package-search-mcp"
+      ]
+    },
+    "github-mcp-server": {
+      "type": "stdio",
+      "command": "npx",
+      "env": {
+        "GITHUB_TOKEN": "${GITHUB_TOKEN}"
+      },
+      "args": [
+        "-y",
+        "@anthropic/github-mcp-server"
+      ]
+    },
+    "searxng": {
+      "type": "http",
+      "url": "http://localhost:8080"
+    },
+    "ollama": {
+      "type": "http",
+      "url": "http://localhost:11434"
+    },
+    "milvus": {
+      "type": "http",
+      "url": "http://localhost:19530"
+    }
+  }
+}
diff --git a/apm.lock.yaml b/apm.lock.yaml
index a78be1249..0c5b77b23 100644
--- a/apm.lock.yaml
+++ b/apm.lock.yaml
@@ -34,9 +34,16 @@ mcp_configs:
   claude-context:
     name: claude-context
     transport: stdio
+    env:
+      EMBEDDING_PROVIDER: Ollama
+      EMBEDDING_MODEL: nomic-embed-text
+      OLLAMA_HOST: http://localhost:11434
+      EMBEDDING_DIMENSION: '768'
+      MILVUS_ADDRESS: http://localhost:19530
+      MILVUS_TOKEN: root:Milvus
     args:
     - -y
-    - '@anthropic/claude-context'
+    - '@zilliz/claude-context-mcp@latest'
     registry: false
     command: npx
   context7:
diff --git a/apm.yml b/apm.yml
index a3014a8e3..82a9576bd 100644
--- a/apm.yml
+++ b/apm.yml
@@ -45,7 +45,14 @@ dependencies:
       command: npx
       args:
         - "-y"
-        - "@anthropic/claude-context"
+        - "@zilliz/claude-context-mcp@latest"
+      env:
+        EMBEDDING_PROVIDER: Ollama
+        EMBEDDING_MODEL: nomic-embed-text
+        OLLAMA_HOST: http://localhost:11434
+        EMBEDDING_DIMENSION: "768"
+        MILVUS_ADDRESS: http://localhost:19530
+        MILVUS_TOKEN: root:Milvus
 
     - name: context7
       registry: false
diff --git a/docs/mcp-user-setup.md b/docs/mcp-user-setup.md
index 68e738a67..6ce08e497 100644
--- a/docs/mcp-user-setup.md
+++ b/docs/mcp-user-setup.md
@@ -35,6 +35,7 @@ Makefile (mcp-up/mcp-down)       # Orchestration: brew + docker compose
 Brewfile                         # System deps: ollama, bun, hugo, rbenv
 ```
 
-- **Stdio MCPs** (claude-context, context7, brave-search, etc.): auto-started by the AI tool's MCP client. No infrastructure needed.
+- **Stdio MCPs** (claude-context via @zilliz/claude-context-mcp, context7, brave-search, etc.): auto-started by the AI tool's MCP client. No infrastructure needed.
+- **claude-context** uses local Milvus (:19530) for vector storage + Ollama (nomic-embed-text) for embeddings.
 - **Ollama**: managed via `brew services`. Started by `brew bundle` (restart_service: true).
 - **SearXNG + Milvus**: Docker services in `compose.mcp.yml`. Started by `make mcp-up`.
diff --git a/opencode.json b/opencode.json
index 333cc299d..36d5e68db 100644
--- a/opencode.json
+++ b/opencode.json
@@ -6,8 +6,16 @@
       "command": [
         "npx",
         "-y",
-        "@anthropic/claude-context"
-      ]
+        "@zilliz/claude-context-mcp@latest"
+      ],
+      "environment": {
+        "EMBEDDING_PROVIDER": "Ollama",
+        "EMBEDDING_MODEL": "nomic-embed-text",
+        "OLLAMA_HOST": "http://localhost:11434",
+        "EMBEDDING_DIMENSION": "768",
+        "MILVUS_ADDRESS": "http://localhost:19530",
+        "MILVUS_TOKEN": "root:Milvus"
+      }
     },
     "context7": {
       "type": "local",