view release on metacpan or  search on metacpan

.claude/rules/api-docker-rules.md  view on Meta::CPAN

# API-Docker House Rules

Apply to every task in this distribution unless explicitly overridden. Bias: caution over
speed on non-trivial work; use judgment on trivial tasks. Loaded automatically at launch
(same priority as `CLAUDE.md`). Subagents get their discipline from the skills
force-loaded via `briefing.skills` — this file is for the orchestrating agent.

## Engineering discipline

1. **Think before coding** — state assumptions; when uncertain, ask rather than guess.
   Push back when a simpler approach exists.
2. **Simplicity first** — minimum code that solves the problem. Nothing speculative.
3. **Surgical changes** — touch only what you must. Match existing style.
4. **Goal-driven execution** — define success criteria, loop until verified.
5. **Surface conflicts, don't average them** — pick one (more recent / more tested), flag
   the other for cleanup. Don't blend.
6. **Read before you write** — `Role::HTTP` is the single seam every resource API and
   entity class hangs off. A change to `_request`'s options, return shape or error
   handling reaches every module in `lib/` and the mock harness at once.
7. **Tests verify intent, not just behavior** — a test that can't fail when the logic
   changes is wrong, and a helper that normalises its input before asserting is that
   test. Reproduce a bug before fixing it; leave the regression behind.
8. **Checkpoint after every significant step** — summarize: done / verified / left.
9. **Match conventions** — conformance > taste. Surface a harmful convention; don't fork
   silently.
10. **Fail loud** — "Done" is wrong if anything was skipped. "Tests pass" is wrong if any
    were skipped — and in this repo a skip is the default failure mode, see below.
11. **A red test is a claim before it is a failure** — before changing code to turn a
    test green, say what the test asserts and whether your fix keeps that claim or
    replaces it. If the claim is wrong, fix the claim and say so.

## Delegation

This rule depends on whether the Agent/Task tool is available to you.

- **You can spawn subagents** (orchestrating main agent): Do NOT touch behavior-relevant
  code yourself — delegate. Your lane: coordinate, inspect, plan, review diffs, run
  tests, manage git, edit non-behavioral docs. When in doubt, delegate. Why: only the
  `api-docker-*` agents get their skills force-loaded via `briefing.skills`; you get no
  briefing and would touch internals with too little context.

  | Task | Agent |
  |---|---|
  | Anything turning on what the daemon does or expects — endpoints, wire formats, filters, registry auth, version gating | `api-docker-engine-worker` |
  | The Perl side — Moo, transport internals, entity classes, refactoring, cpanfile | `api-docker-worker` (default) |
  | Write/extend tests, add fixtures | `api-docker-test-writer` |
  | The generated type model, the `API::Docker::Type` DSL, the drift checker, `spec/` | `api-docker-type-writer` |
  | Pre-release audit | `api-docker-release-checker` |
  | POD and README | `api-docker-doc-writer` |

  The two workers split by *question*, not by file: "what does the engine answer here?"
  is the engine-worker's, "how is this distribution built?" is the plain worker's. Only
  the engine-worker carries the Engine API reference — the other one guessing at daemon
  behavior is how a wrong assumption gets cemented.

- **You cannot spawn subagents** (you ARE an `api-docker-*` agent): the delegation lock
  does not apply to you — implement, refactor, debug, and test per these rules.

Behavior-relevant = the HTTP transport and everything it returns, the resource API method
surface, entity wrappers, request/response encoding, error handling, `cpanfile`, and
tests. Pure prose docs and `Changes` notes are not.

## Parallel fan-out — isolate the working tree

Subagents share one working tree with the orchestrator and with each other. A global git
command in one reaches all of them, so:

- **A subagent never mutates git** — no `stash`, `reset`, `checkout -- <path>`, `clean`,
  `add` or `commit`. The orchestrator owns git and commits. Say so in every subagent
  prompt, but do not rely on the prompt alone: a subagent's `git stash`/`reset`/`checkout`
  has thrown away another agent's uncommitted work three times (k111) even when the prompt
  forbade it.
- **When two or more code-touching agents run at once, isolate them.** Launch each with
  `isolation: "worktree"` so a stray git command in one cannot reach another's tree, or run
  them sequentially in the shared tree. Never fan out parallel code-touching agents into the
  same working tree without isolation.
- **A worktree may branch from a stale base.** Integrate its result by the diff
  (`git diff <merge-base> <branch> -- <files>` piped to `git apply`, or a cherry-pick),
  never by `git checkout <branch> -- <file>` for a file the main tree has since changed —
  that reverts the main tree to the stale copy. Check the merge-base against what main