agentleFS
ChindanaiNaKubChindanaiNaKub/openthought/.claude/skills/ship-check/SKILL.md

ship-check

Skill0 starsChanged 4 months agoMITView on GitHub

Gate a diff against the intent it claims to serve before committing, pushing, or opening a PR — check that changed files trace to an artifact, that the story's acceptance criteria are actually met, and that the change didn't quietly widen beyond what was asked. Use before any commit or PR, when the user asks "is this ready", "does this match what I asked for", or after finishing a chunk of agent-written work.

19 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use review.

Other files people use alongside this one.

Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
  • SentrySkill

    review

    Fixture review skill.

    Also in 3 projects that use this

The file

We split this skill into the rules it gives an agent

2 rules. 0 are common, 1 is distinctive, and 1 is very specific to this project.

  • 0

    Common

    Most of the 18 edited versions we read keep these.

  • 1

    Distinctive

    This author's own choices. Worth a look.

  • 1

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Read the diff (git diff, plus git diff --cached and any untracked files).

    line 45
  2. Very specific:

    Do not invent an artifact to satisfy the gate.

    line 82

“Common” is measured: kept by at least 6 in 10 of the 18 edited versions of this file we could read. “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
SKILL.md

Ship check

The moment before a commit is the last cheap point to notice that the diff answers a question nobody asked. This gate runs then.

coherence ship                    # gate the working tree
coherence ship --story story-x    # against a specific artifact
coherence ship --base main        # diff against a ref instead
coherence ship --json             # structured, for your own reasoning

(Commands assume coherence on PATH; a repo may vendor it at bin/coherence instead. Use whichever exists.)

Exit 1 means something blocking. It reports rather than locks. A solo builder can always override, and should be told what they are overriding instead of stopped.

What it does mechanically

  1. Collects changed files (working tree, staged, and untracked).
  2. Maps each onto artifacts via their code: globs.
  3. Resolves what the diff claims to serve, from --story or coherence focus.
  4. Reports:
    • untraced-change: changed files no artifact owns. Blocks when nothing is in focus at all. Warns when a story is focused but the files fall outside its globs, because that is usually scope creep.
    • stale-premise: you are shipping under an ancestor that is dropped or done. Blocks. The reason for this work may be gone.
    • no-criteria: the focused story has no acceptance criteria, so there is nothing to judge the diff against.
    • coverage-gap: the story serves no problem or user.
  5. Prints the acceptance criteria.

What you do with it

The mechanical part stops at printing the criteria. You judge the diff against them, and that judgment is the actual gate.

Read the diff (git diff, plus git diff --cached and any untracked files). Then for each criterion, decide met, unclear, or unmet, and say which line of the diff decides it. "Unclear" is a real and useful answer. Use it when the criterion is not written precisely enough to check, and say so, because that is a finding about the artifact rather than about the code.

Then answer the two questions the criteria do not cover.

Does the diff do less than it claims? Criteria ticked in the artifact but not actually satisfied by the code. This is the common agent failure: the work is plausible and incomplete.

Does the diff do more than it claims? Files, refactors, dependencies, or abstractions the story never asked for. This is the other common agent failure, and the one nobody catches, because extra work looks like diligence. A rename that touched forty files, a helper generalised for a second caller that does not exist, error handling for a case the story excluded: each is a separate decision the user never made. Name them individually.

Reporting

Be specific and short. Something like:

story-magic-link, 3 of 4 criteria met.

  • link expires after 15 minutes, src/auth/token.py:41
  • single use, enforced by the used_at column
  • rate limited per address, src/auth/routes.py:88
  • failures log without leaking the token: token.py:96 logs the full URL, token included.

Also outside this story: src/db/pool.py was rewritten to use a connection pool. Nothing asked for that. It may be right, but it is its own decision.

Then state plainly whether it is ready. If it is, say so without hedging.

When nothing is in focus

Do not invent an artifact to satisfy the gate. Ask what this work serves, or search intent/ for something that already covers it and offer to focus that. If the work genuinely serves nothing in the graph, that is the finding. Either the graph is missing something worth writing down, or the work is worth questioning.

Related: intent-graph for authoring, coherence-check for the standing checks.

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.