chore: modernize coding-agent harness - #65
Merged
Conversation
.claude/skills is now a symlink into .agents/skills (previously the reverse), so agent-neutral tooling reads the real path and Claude Code follows the link. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
- create-pr, grill-me, grill-with-docs, handoff, tdd, improve-codebase-architecture, updating-deps(-auto) updated to current best-practice versions; superseded tdd/architecture companion docs removed (their content now lives in the codebase-design skill) - ADR/CONTEXT format references moved to the domain-modeling skill - code-reviewer agent is now a thin wrapper over the shared code-review-guide skill with a cross-model second opinion - product-advisor gains a read-the-repo-first analysis process - backend rules: Testing + Dependencies sections; fixed stale claim that the project has no pgvector usage - frontend rules: Data fetching section (TanStack wrappers, generated client is read-only) - per-skill agents/openai.yaml shims for cross-agent discovery Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
New skills: grilling, domain-modeling, codebase-design, code-review-guide (shared review methodology used by both harnesses' code-reviewer agents), writing-great-skills, implement, to-spec, to-tickets, and setup-matt-pocock-skills. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
hooks.json wires the same three guard/lint hooks (byte-identical copies under .codex/hooks/), and agents/*.toml port the Claude subagents so both CLIs share one methodology. The code-reviewer's cross-model second opinion calls the other CLI from each side. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The test suite (159 tests) was not running in CI at all. New test job runs it DB-free with a 40% coverage floor (baseline ~43%, to be ratcheted up). eval.yml runs the real-LLM pydantic-evals suite on manual dispatch only, needing the OPENAI_API_KEY repo secret. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Points agents at the canonical skills directory, the Codex mirror sync rule, the shared code-review methodology, and the skill-authoring and spec-driven workflows. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Brings the repo's coding-agent harness up to current best practices: a richer skill suite, dual-agent (Claude Code + Codex) parity, a shared code-review methodology with cross-model second opinions, and CI test/eval jobs that were missing entirely.
Skills
.agents/skills/with.claude/skillsas a symlink (previously reversed), so agent-neutral tooling reads the real path. History preserved viagit mv.grilling,domain-modeling,codebase-design,code-review-guide,writing-great-skills,implement,to-spec,to-tickets,setup-matt-pocock-skills.create-pr,grill-me,grill-with-docs,handoff,tdd,improve-codebase-architecture,updating-deps,updating-deps-auto); superseded companion docs removed in favor of thecodebase-designvocabulary skill.agents/openai.yamlshim for Codex discovery.Codex parity (
.codex/)hooks.jsonwires the same three guard/lint hooks (byte-identical copies).agents/*.tomlport the three Claude subagents; thecode-revieweron each side delegates to the sharedcode-review-guideskill and gets a cross-model second opinion from the other CLI (with a SOLO recursion guard).CI
testjob — the 159-test suite was not running in CI at all. Runs DB-free with a 40% coverage floor (baseline ~43%, to be ratcheted up); addspytest-cov.eval.yml— real-LLM pydantic-evals on manual dispatch only.OPENAI_API_KEYrepo secret before first dispatch.Rules & agents
code-revieweris now a thin wrapper overcode-review-guide;product-advisorgains a read-the-repo-first analysis process.TrackAudioEmbedding's HNSW index).AGENTS.mdgains an Agent Harness section documenting the symlink/mirror topology.All project-specific guidance (Modal rules, DEAP setup, pydantic-ai tool-error convention, ports, architecture map) is preserved unchanged.
Verification
pre-commit run --all-filesgreen (ruff, ruff-format, mypy --strict).claude/hooks↔.codex/hooksverified byte-identicalgit log --follow)🤖 Generated with Claude Code