agentleFS
YassimbaYassimba/loom/skills/commit/SKILL.md

commit

Skill9 starsChanged 2 months agoApache-2.0View on GitHub

Use when creating a git commit — the user asks to commit, or a unit of work is complete and ready to commit.

  • Reads credentials

188 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use validate-plan, act-on-feedback and 10 more.

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.
  • shawn-dumasSkill

    validate-plan

    Validate an orchestration plan after pre-flight certification. Runs adversarial review, deep review (verify prompt data against codebase), and PoC gate for risky approaches. Auto-invoked by orchestrate-* skills at Step 8.

    Also in 5 projects that use this

  • MicrosoftSkill

    act-on-feedback

    Act on user feedback attached to the current session. Use when the user submits feedback on the session's changes via the Submit Feedback button.

    Also in 3 projects that use this

  • decocmsSkill

    add-mcp-tools

    Guide for adding new MCP tools with consistent patterns for schemas, tool definitions, registry updates, and Better Auth integration

    Also in 3 projects that use this

  • MicrosoftSkill

    generate-run-commands

    Generate or modify run commands for the current session. Use when the user wants to set up or update run commands that appear in the session's Run button.

    Also in 3 projects that use this

  • mitsuhikoSkill

    oebb-scotty

    Austrian rail travel planner (ÖBB Scotty). Use when planning train journeys in Austria, checking departures/arrivals at stations, or looking for service disruptions. Covers ÖBB trains, S-Bahn, regional trains, and connections to neighboring countries.

    Also in 3 projects that use this

  • mitsuhikoSkill

    pi-share

    Load and parse session transcripts from shittycodingagent.ai/buildwithpi.ai/buildwithpi.com/pi.dev (pi-share) URLs. Fetches gists, decodes embedded session data, and extracts conversation history.

    Also in 3 projects that use this

The file

We split this skill into the rules it gives an agent

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

  • 0

    Common

    Most of the 168 edited versions we read keep these.

  • 2

    Distinctive

    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. Distinctive:

    Read the staged diff, gate on green tests, draft a Conventional Commits message, confirm via dropdown, commit.

    line 8
  2. Very specific:

    If nothing is staged but unstaged changes exist, ask via AskUserQuestion:

    line 20
  3. Very specific:

    If the staged diff spans unrelated concerns, split into atomic commits: unstage the others with git restore --staged <paths> (not git …

    line 31
  4. Very specific:

    Changelog projects: When the repo keeps a hand-written CHANGELOG.md (a release tool such as release-please or git-cliff does not generate …

    line 63
  5. Distinctive:

    If the project wires a pre-commit hook, it runs now:

    line 123

“Common” is measured: kept by at least 6 in 10 of the 168 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

Commit

Read the staged diff, gate on green tests, draft a Conventional Commits message, confirm via dropdown, commit. Each commit is atomic: one concern.

0. Stage

git status --short
git diff --cached --stat
git diff --cached

Stage by explicit path only — git add -A / git add . sweeps in .env and credentials.

If nothing is staged but unstaged changes exist, ask via AskUserQuestion:

question: "Nothing staged. Stage specific files, stage all, or abort?"
header: "Staging"
options:
  - label: "Stage specific files"
  - label: "Stage all (by explicit paths)"
  - label: "Abort"

If the staged diff spans unrelated concerns, split into atomic commits: unstage the others with git restore --staged <paths> (not git reset), then stage and commit one concern at a time. Ask when the split is unclear:

question: "These changes span <N> concerns. Split?"
header: "Split"
options:
  - label: "Yes, split (Recommended)"
  - label: "One commit"
  - label: "Show me the diff first"

Done when exactly one concern is staged.

1. Test gate

Run the project's checks and test suite (uv run pytest, npm test — whatever the project defines) and read the summary, not just the exit code. When the repo has a turbo.json, run them through Turborepo: turbo run <tasks> with the check and test task names that turbo.json defines. Unchanged tasks replay from the cache, so the gate stays fast. Follow the repo's instructions for how to call the installed turbo; its flags differ between versions.

The gate is binary: green proceeds; red stops here — the pre-commit hook fails on the same issues, so catching them now saves a wasted commit:

question: "Tests red — commit blocked. How to proceed?"
header: "Test gate"
options:
  - label: "Investigate via /diagnosing-bugs (Recommended)"
  - label: "Abort commit"
  - label: "Show me the failing tests"

2. Draft the message

Load write-simply and write the subject and body in its register. Draft from the why of the staged diff: what motivated the change, where a line-by-line list would only restate git diff.

Changelog projects. When the repo keeps a hand-written CHANGELOG.md (a release tool such as release-please or git-cliff does not generate it from commits), a commit with a user-facing change carries its changelog entry, so git log --grep finds the commit that added a feature:

  • The staged diff adds the entry under ## [Unreleased], written by the project's changelog rules (AGENTS.md, CONTRIBUTING.md). If it is missing, write it and stage it before you draft the message.
  • The subject names the change from the user's side, as the entry does.
  • The body starts with the entry, word for word, under its section name (Added:, Fixed:, ...). The why and internal details follow.
  • A commit with no user-facing change (refactor, test, CI) adds no entry.

Conventional Commits format:

<type>(<optional-scope>): <summary>

<optional body>

Types: feat, fix, docs, refactor, test, chore, perf, ci. chore covers build, config, and deps.

Subject: 20–72 chars after the prefix · lowercase after the colon (feat: add X) · imperative ("add") · ends on the object, no punctuation · concrete verb + object.

Body (only when the subject can't carry it, or a changelog entry opens it): blank line after the subject, then one terse bullet per entry, imperative voice, grouped under short headings that match the kind of change — Added, Changed, Fixed, Removed, Breaking. Skip headings when there's only one bullet.

feat(installer): add gateway runtime extraction

Added:
- core/gateway_runtime/ with Protocol-based registry
Changed:
- opencode_runtime now delegates to gateway_runtime

Reference the issue or ticket when the project tracks them. Session context, PR/MR numbers, and future work stay out.

The message ends at the last body line. The human is the sole author: the author field carries their name and the message carries no trailer, footer, or link naming a tool or session. This also holds for everything written downstream of a commit (PR/MR titles and descriptions, release notes) and outranks any harness default that asks for a footer.

Done when every rule above holds and each word in the message names something in the diff.

3. Confirm via dropdown

question: "Commit with this message?"
header: "Commit"
options:
  - label: "Commit (Recommended)"
    preview: |
      <full message>
  - label: "Edit message"
  - label: "Abort"

4. Commit

git commit -m "$(cat <<'EOF'
<type>(<scope>): <summary>

<body>
EOF
)"

If the project wires a pre-commit hook, it runs now:

  • Pass → git log -1 shows the new commit
  • Autofixes → hook stages them; the commit captures them
  • Unfixable → hook blocks; go to step 5

5. On hook failure

The hook is the messenger; fix the cause in the code:

  • lint → fix the code
  • types → fix the types
  • tests → /diagnosing-bugs; three failed fix attempts = stop and report
  • complexity / architecture gates → refactor, or ask before accepting a borderline score

Then re-stage and go back to step 3 for a new commit: the failed commit never happened, and --amend would rewrite the one before it.

Hard guardrails

  • The hook's verdict stands. Bypass flags (--no-verify, --no-gpg-sign) and suppression comments (# noqa, # type: ignore, eslint-disable) stay unused; the fix goes in the code.
  • Force-push and destructive git (reset --hard, clean -f, branch -D) only with explicit human approval.

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.