feat(mcp): Docker MCP infrastructure — milvus, searxng, healthchecks, and claude-context migration

.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

.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

.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

.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

.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

.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

.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

.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

.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

.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

.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

.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

.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

.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

.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

.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