agentleFS
prototypoprototypo/codequalityexplorer/CLAUDE.md

codequalityexplorer

CLAUDE.md2 starsChanged 46 days agoMITView on GitHub

The project-manager uses the standard agents by default. The heavy variants run on a larger model and cost more per call, so use them only where these rules say. Use developer-heavy instead of developer when a task: - changes concurrency, locking, transactions, or background job coordination - is a refactor that changes behaviour across more than one module or public interface - changes authentication, authorisation, session or token handling, or cryptography - touches any path listed under "Security-critical paths" below…

The file

We split this CLAUDE.md into the rules it gives an agent

5 rules. About 0 are common, 2 are distinctive, and 3 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈2

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 3

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Use developer-heavy instead of developer when a task:

    line 7
  2. Distinctive:

    Do not use it for large but mechanical changes (renames, boilerplate, CRUD endpoints that follow an existing pattern, test data).

    line 14
  3. Very specific:

    Use security-reviewer-heavy instead of security-reviewer when the diff:

    line 18
  4. Very specific:

    If developer-heavy fails the same gate twice for the same finding, stop and report to the user.

    line 35
  5. Distinctive:

    Every request for a new feature MUST run this gate sequence, in order.

    line 40

≈ “Common” is an estimate: at least 1 in 5 of the 26,980 CLAUDE.md files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
CLAUDE.md

Agent routing

The project-manager uses the standard agents by default. The heavy variants run on a larger model and cost more per call, so use them only where these rules say.

developer-heavy

Use developer-heavy instead of developer when a task:

  • changes concurrency, locking, transactions, or background job coordination
  • is a refactor that changes behaviour across more than one module or public interface
  • changes authentication, authorisation, session or token handling, or cryptography
  • touches any path listed under "Security-critical paths" below
  • is an escalation (the standard developer failed the same gate twice for the same finding)

Do not use it for large but mechanical changes (renames, boilerplate, CRUD endpoints that follow an existing pattern, test data).

security-reviewer-heavy

Use security-reviewer-heavy instead of security-reviewer when the diff:

  • touches any path listed under "Security-critical paths" below, regardless of which developer wrote it
  • adds or changes an endpoint, route, or handler that enforces access control
  • adds or changes cryptography, token generation or verification, or secret handling
  • adds a new third-party dependency that runs at a trust boundary (parsers, auth libraries, crypto libraries)

Security-critical paths

<!-- Replace with your project's actual paths. -->
  • src/auth/
  • src/permissions/
  • src/crypto/
  • src/api/middleware/
  • Any database migration that changes roles, permissions, or tenant isolation

Limits

  • If developer-heavy fails the same gate twice for the same finding, stop and report to the user. Do not keep looping.
  • The code-reviewer, tester, and documenter have no heavy variants.

Feature workflow (mandatory)

Every request for a new feature MUST run this gate sequence, in order. Do not skip a gate unless the user explicitly waives it for that request.

  1. Grill — invoke mattpocock-skills:grilling to stress-test the idea and requirements before any design work. If grilling surfaces new domain terms or an architectural decision, run mattpocock-skills:domain-modeling (CONTEXT.md / ADR) before planning. If a design question can only be answered by trying it, run mattpocock-skills:prototype; if it needs external facts, mattpocock-skills:research.
  2. Plan — enter plan mode; get the plan approved before touching code.
  3. Delegate — hand the approved plan to the project-manager agent. The PM runs the pipeline: developer implements via mattpocock-skills:tdd (test-first; red-green-refactor) → code-reviewer gate → security-reviewer gate; failed gates route fixes back to the developer until both PASS.
  4. Review — run mattpocock-skills:code-review against the merge-base (Standards axis + Spec axis vs the approved plan); route confirmed findings back through the project-manager.
  5. Commit — tester runs the full test suite(s) (both repos if the change touches both) and commits per repo only when everything passes; then documenter updates docs (including docs/integration-contract.md if the handoff changed).
  6. Verify — invoke superpowers:verification-before-completion: show fresh test output before claiming the feature is done.

Bug workflow

Bugs skip Grill/Plan. Diagnose first via mattpocock-skills:diagnosing-bugs (reproduce before you touch code), then enter the pipeline at Delegate: the developer turns the reproduction into a failing test, fixes it, and the same gates (code-reviewer → security-reviewer → tester → documenter) apply.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.