agentleFS
blueclover22blueclover22/mw_agent_kit/skills/commit/SKILL.md

commit

Skill1 starsChanged 8 days agoMITView on GitHub

Use when the user explicitly asks to commit the current changes. Push, amend, and rebase require separate explicit requests.

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, 5 are distinctive, and 0 are very specific to this project.

  • 0

    Common

    Most of the 168 edited versions we read keep these.

  • 5

    Distinctive

    This author's own choices. Worth a look.

  • 0

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Not verified, and the project declares none → report this gap and the remaining verification blind spot to the user, and get explicit …

    line 22
  2. Distinctive:

    Follow the repo's convention (recent git log style, commitlint) only if it is consistent and informative.

    line 32
  3. Distinctive:

    Style — concise and declarative, never narrative: Every line carries a fact, not prose about the work:

    line 39
  4. Distinctive:

    No commit with a failing or skipped verification gate without the user's explicit confirmation.

    line 83
  5. Distinctive:

    No file edits by this skill — staging/unstaging and committing only, the commit-message temp file being the sole exception.

    line 84

“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

mak:commit — Work Wrap-Up and Commit

Running this skill is the user's explicit "commit" instruction — for the commit only.

Procedure

1. Inspect

git status + git diff (staged / unstaged / untracked). Nothing to commit → report and stop. Identify which task the changes belong to.

2. Gates

Results are reported once, in the §5 report. Surface immediately only findings that need a user decision (split proposal, unverified proceed, suspected secret).

  • Verification — confirm the changes were verified (mak:verify-checklist results or the project's verify commands).
    • Not verified, and the project declares verification commands (per the mak:verify-checklist §Verification Order priority rule, which owns what counts as declared) → run those commands as documented, with no auto-fix or formatter flags. Build/type check, lint, and tests are common examples, but run whatever the project declares — a docs-only project may declare a manifest validator and consistency greps instead. Then re-check git status since builds and tests produce artifacts. On failure, either get explicit confirmation to proceed unverified, or stop and route the fix through the normal flow before re-entering this skill.
    • Not verified, and the project declares none → report this gap and the remaining verification blind spot to the user, and get explicit confirmation before proceeding unverified.
  • Precise changes — one purpose per commit. Unrelated changes mixed in → propose a split or leave them unstaged; never bundle silently.
  • Hygiene — exclude temp/junk files, logs, large binaries, and secrets. Never commit a suspected secret without asking, even if staged.

3. Stage deliberately

Stage the reviewed files explicitly; no blind git add -A.

4. Message

  • Follow the repo's convention (recent git log style, commitlint) only if it is consistent and informative. Placeholder history ("update", "fix", "wip"), or too thin/mixed to establish a deliberate convention, is not one — use the fallback and note briefly why.
  • Fallback and quality floor: Conventional Commits — <type>(<scope>)?: <subject>, imperative subject ≤ 72 chars. A repo convention may override the shape, never the informativeness.
  • Types, unless the repo's own set differs: feat fix test docs refactor (no behavior change — if behavior moved, it is feat/fix) perf (no functional change) style (code formatting, never UI styling) build (build system, dependencies) ci (CI config) chore (maintenance fitting no other type — not a bucket for changes left unclassified) revert (name the reverted commit in the body). Do not invent one outside the set in use.
  • Type = the change's primary intent, not its file type. One type per commit — two genuinely applying is a split signal (back to §2); docs/tests belonging to a feat/fix ship inside it. Scope optional (module/domain; reuse scopes from git log). Breaking change: ! + BREAKING CHANGE: body line.
  • mak outputs: design doc / roadmap / doc-set → docs; a verify-caught fix before commit → part of the original feat/fix; kit/config housekeeping → chore.
  • Write the message in the user's conversation language (or the project's documented language policy).

Style — concise and declarative, never narrative. Every line carries a fact, not prose about the work:

  • Subject — the outcome ("remove the cross-layer duplication"), not the mechanics ("move a file"). Imperative or noun phrase, no trailing period.
  • Body — 2–6 bullets, one line per bullet, one verifiable fact each with counts/names. No paragraphs, no multi-sentence bullets, no prose connecting them.
  • Cut, always — "this commit …" openers, background/motivation essays, restating the diff file by file, closing summaries, hedges ("slightly", "some"), and polite or past-tense sentence endings (in languages with such endings — e.g. Korean — use noun-phrase endings instead).
  • Keep — preserved invariants (contract unchanged, tests moved along), what intentionally stayed put, doc-sync section names.
  • Scale the body to the diff — a tiny change takes the subject alone or at most one bullet; multi-bullet bodies are for diffs with multiple facts worth recording.

The terseness rule holds in every language; only the wording changes. Example (imitate the structure, in the user's language):

refactor(order): remove OrderStatus duplication via the shared domain layer

- OrderStatus: features/checkout/model -> domain/order (serialization contract unchanged, 2 unit tests moved)
- payment/history references switched to the domain path — 1 cross-feature import exception removed
- OrderStatusDto stays in features/checkout (API response shape)
- Docs sync: architecture guide (§module tree, §exception list), glossary

5. Commit and report

Pre-commit self-check: the message body (logical view) and the staged files (physical view) must agree — a bullet with no supporting file, or a file no bullet accounts for, is a stop signal; resolve it before committing.

Never pass the message as a shell argument. Write it to a temp file outside the repository (session scratchpad or OS temp dir), run git commit -F <file>, then delete the file. -m quoting rules differ per shell and silently corrupt multi-line messages — a PowerShell here-string (@'…'@) executed under bash commits @ as the subject. A temp file inside the working tree is not an option: it shows up as untracked and can be swept into a later commit.

Then report in the user's language (one commit per purpose; more if a split was agreed in §2):

## Commit Result
- <short hash> — <the committed message: subject and body>
- Gates: verification ✅/⚠️ · precise-changes ✅/⚠️ · hygiene ✅/⚠️

| File | +/− | What changed |
| :--- | :--- | :--- |

- Left uncommitted (intentionally): <files + reason, or "none">
- Not pushed — push runs only on the user's explicit request

Cluster long file lists into logical groups, but every file appears.

Prohibitions

  • No push / --amend / rebase / reset / force operations / tags on your own — each only on the user's explicit request for that operation.
  • No commit with a failing or skipped verification gate without the user's explicit confirmation.
  • No file edits by this skill — staging/unstaging and committing only, the commit-message temp file being the sole exception; fixes go back through the normal flow. Artifacts incidentally produced by verification (caches, build output) are never staged or committed.

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.

Reviews can't be read right now.

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.