agentleFS
phmatrayphmatray/ai-migration-kit/.claude/CLAUDE.md

ai-migration-kit

CLAUDE.md4 starsChanged 59 days agoMITView on GitHub

This repo ships as a Claude Code plugin (bash + python + markdown skills, no build step). The per-repo profile at .claude/skills/repo-profile.md is the single source of commit identity, build/test commands, CI gates, labels and conflict hot-spots — read it first. The kit is what is being dogfooded here, so its own skills come first. Which one, for what: the routing table in AGENTS.md — its one home since #525, where every host reads it — imported here so a…

1 other project uses a version of this.

Adopted by

Most recently taken up by

The file

Files
CLAUDE.md

Working on tagout

This repo ships as a Claude Code plugin (bash + python + markdown skills, no build step). The per-repo profile at .claude/skills/repo-profile.md is the single source of commit identity, build/test commands, CI gates, labels and conflict hot-spots — read it first.

Routing

The kit is what is being dogfooded here, so its own skills come first. Which one, for what: the routing table in AGENTS.md — its one home since #525, where every host reads it — imported here so a session in this repository sees it whether or not the plugin is installed:

@../AGENTS.md

Where each concern lives

  • Prerequisites (runtime) → requirements.json, never a hard-coded list; scripts/preflight.sh reads it.
  • Control-flow decisions → decisions/registry.json + docs/decisions.md — one id, one program, one home.
  • Architectural decisions → the profile's ## ADRs section (.claude/skills/repo-profile.md) — docs/adr/.
  • Domain language → CONTEXT.md.
  • Trigger contracts → evals/<skill>-trigger-eval.json, the one home (#331); structure checked by tests/skills/check-frontmatter.py.
  • Shared test preamble → tests/_lib.sh (local rc=$? must be the first statement in its trap).
  • Shared skill procedures → skills/_shared/.
  • Kit backlog (YAGNI debts) → docs/backlog.md, hand-edited, read by scripts/followups.py.
  • The two plugin directories → generated copies, never hand-edited (ADR 0017): edit the source under skills/, scripts/, hooks/…, then python3 scripts/host-adapters.py build and commit both halves; check refuses a drifted copy, an orphan, a symlink (#619) or a mode mismatch.

Adding a skill

  • The folder name obeys one of the two naming rules — verb-object, or <family>-<role> for a family member. The name is the identifier six times over, and renaming one is a breaking change, so it only happens in a major: ADR 0012.
  • Frontmatter carries no version key — tests/skills/check-frontmatter.py.
  • A trigger contract exists for it — evals/<skill>-trigger-eval.json, listed in evals/run_all.py and evals/trigger_eval.py.
  • A golden suite is wired into CI — scripts/ci-wiring-check.py.
  • A script that makes a decision is registered in decisions/registry.json, or named in that file's not_decisions map with a one-line reason.
  • The skill is linked from README.md — tests/skills/test.sh.
  • The PR title is releasable — scripts/release-title-gate.sh.

The guard convention

A destructive operation gets a guard script under skills/<skill>/scripts/, a golden test that exercises its refusal path (not just its happy one), and a CI step that runs it — never the raw command. See the README's "Hardening a destructive operation".

Releases

Squash-merge only; the PR title is the release commit release-please parses (scripts/release-title-gate.sh gates it). Never bump plugins/tagout/.claude-plugin/plugin.json (and its tagout-migrate twin) by hand.

Commit identity

The profile's author line (.claude/skills/repo-profile.md) is canonical; this file sets none.

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.