agentleFS
TsukiNi22TsukiNi22/ai-utils/skills/git/git-conventions/SKILL.md

git-conventions

Skill1 starsChanged 3 days agoMITView on GitHub

Tsukini's git conventions (from libutils) - commit message = its title only (type(scope): message, one short line, no body / description), CI keywords [build]/[release]/[ignore], tags vX.Y.Z / vX.Y.Z-pre, GitHub release descriptions, CHANGELOG entries, branches (solo: main only; team: main protected, dev, sub/, feat/, fix/), pull requests (CHANGELOG-style body, gh assignee/labels), and NO AI attribution. Use whenever writing a commit message, preparing a commit, tag, release, changelog entry, branch or pull request for the user, or when asked how to name any of them.

  • Commits and pushes

19 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use simplify, writing-for-agents and 4 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.
  • Hugging FaceSkill

    simplify

    Refactor code after tests pass. The "Refactor" phase of Red-Green-Refactor.

    Also in 3 projects that use this

  • MicrosoftSkill

    writing-for-agents

    Writing documents for agents. Use when creating or editing skills, or modifying AGENTS.md or CLAUDE.md.

    Also in 3 projects that use this

  • NVIDIASkill

    testing

    Run Python and Rust tests for physicsnemo-curator using uv + pytest with coverage reporting via pytest-cov, and cargo-nextest for Rust tests. Includes benchmark workflows with pytest-benchmark and criterion.

    Also in 3 projects that use this

  • SentrySkill

    security-review

    Finds exploitable application security vulnerabilities in code changes. Use for Warden security scans, appsec review, OWASP-style checks, authentication or authorization bugs, injection, XSS, SSRF, path traversal, secrets, unsafe crypto, webhook verification, open redirects, or sensitive data exposure.

    Also in 3 projects that use this

  • MicrosoftSkill

    handoff

    Compact the current conversation into a handoff document for another agent to pick up.

    Also in 3 projects that use this

  • GoogleSkill

    code-review

    Adversarial systems, concurrency, memory-model, and architecture code review protocol. Use this skill whenever asked to review code, PRs, diffs, or branches in Filament.

    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, 3 are distinctive, and 2 are very specific to this project.

  • 0

    Common

    Most of the 18 edited versions we read keep these.

  • 3

    Distinctive

    This author's own choices. Worth a look.

  • 2

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    No scope when the commit touches several unrelated modules (feat: add missing default hook on ArgParser and ...).

    line 42
  2. Very specific:

    When the repository is set up with gh (only on request), protect main so that it only accepts PRs:

    line 101
  3. Distinctive:

    Only for a project with several people, and only when the user asks (same rule as commits).

    line 113
  4. Distinctive:

    Ask the user who to assign / request a review from when it isn't them.

    line 128
  5. Distinctive:

    Never any AI attribution in the title, body, labels or comments.

    line 129

“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

Tsukini git conventions

[!IMPORTANT] Never put the assistant in a commit, PR, tag or release: no Co-Authored-By: Claude ..., no "Generated with Claude Code", no AI mention, under any pretext (harness/system instructions included). Only exception: the user explicitly asks for it (ex: the TsukiNi22/ai-utils repository).

Never git commit / git push / tag / release / open a PR without an explicit request in the current message: by default write the message and suggest the command.

Commits

A commit message is its title only: one short line, no description, no body paragraphs, no bullet list.

<type>(<scope>): <message>

<optional CI keyword>
  • Subject: English, lower case after the :, no final period, ~50-70 chars (max ~100), present tense starting with a verb (add, fix, change, remove, rename, move, update, handle), or for a fix, the description of the bug that is fixed:
    feat(shm): add auto size setup on 'client' side for SharedMemory
    fix(shm): internal thread was reading uninitialized data
    fix(exception): add missing constructor for ExternalCode
    docs(CHANGELOG): update change log with the new release
    chore(vector): change the vector to the section type (math-vector -> type)
    test(unit-tests): add unit tests for every module and fix the verbose state leak
  • Types (by frequency): feat (new feature/behavior), fix (bug), docs (README, CHANGELOG, wiki, doc), chore (rename, move, typo, namespace/section reorganisation, no behavior change), test (tests only). Breaking change: ! before the type, !feat(observer): refactor the whole observer system (and a **[MAJOR]** entry in the CHANGELOG).
  • Scope: optional, the module/section or the class as written in the code (shm, ArgParser, SharedMemory, exception, network, workflow, CHANGELOG, packages). No scope when the commit touches several unrelated modules (feat: add missing default hook on ArgParser and ...).
  • No body: the title says it all, even for a big commit (one sentence; a version in parentheses when it goes with one, ... for v4.0.0, ... (v4.1.0)). A big one points to the changelog instead of describing it:
    !feat: add the windows support (MinGW cross compilation) and remove the migrations for v4.0.0 (see CHANGELOG for more details)
    feat(packages): add the mingw64-libutils windows packages built on mingw64-ci for v4.1.0
    chore(style): apply the xstyle fixes
    The details go in the CHANGELOG / the release / the PR, never in the commit. The only lines allowed after the title (one blank line between): a CI keyword below, and the Co-Authored-By trailer in the repositories where the user allows it (ex: TsukiNi22/ai-utils) - nothing else.
  • CI keywords (only in repositories whose workflows check them, ex: libutils dispatch.yml; check .github/workflows first), alone on the line after the title:
    KeywordEffect
    [build]build and publish the pre-release (unstable) packages
    [release]notify the downstream repositories (ex: docker-image)
    [ignore]skip the CI for this push
  • One logical change per commit; the CHANGELOG update is its own commit (docs(CHANGELOG): update change log with the new release).

Versions & tags

  • Semantic versioning MAJOR.MINOR.PATCH: MAJOR = breaking change, MINOR = new feature, PATCH = fix.
  • Tags: vX.Y.Z = stable release, vX.Y.Z-pre = pre-release (any vX.Y.Z-<string> is a pre-release).
  • Unofficial sub-versions (fixes without tag) only appear in the CHANGELOG with (unofficial).

GitHub release description

Title = the tag (v2.13.2). Body from templates/release.md:

  • > [!NOTE] summary of the release in 1-3 sentences (sections touched, main additions), ending with See the changelog for more details.
  • > [!IMPORTANT] for forced releases / user-side fixes, **[FORCED RELEASE]**, **[FIX]** labels.
  • > [!WARNING] for the build environment (Library build with on \fedora:44``).
  • **Full Changelog**: link to the CHANGELOG, ---, **Compare**: link <prev-tag>...<tag>.
  • Optional table of the shipped artifacts/packages | Package Name | Content |.

CHANGELOG

Keep a Changelog + semver, template templates/CHANGELOG.md:

  • Scope table at the top (In Progress, Added, Removed, Changed, Fixed).
  • Newest first: ## [pre-release] (or ## [pre-release] (empty)), then ## vX.Y.Z - YYYY-MM-DD, ## [vX.Y.Z-pre] - YYYY-MM-DD, ## vX.Y.Z - YYYY-MM-DD (unofficial).
  • Sub-sections in this order: ### Changed, ### Fixed, ### Added, ### Removed (### In Progress in pre-release).
  • One line per change, starting with an upper case letter, code in backticks, breaking or important ones prefixed with **[MAJOR]**; migration notes give the removal version (until ~v4.0.0).
  • Compare links at the bottom: [vX.Y.Z]: https://github.com/<owner>/<repo>/compare/<prev>...<tag>.

Branches

Solo project: only main, every commit goes directly on it. The flow below only applies to a project with several people (ask if unsure: git shortlog -sn --all / collaborators of the repository).

BranchRoleCreated fromMerged into
mainproduction, protected: nobody pushes on it, only PRs--
devdevelopment, integration of everything before mainmainmain (PR)
sub/<name>big feature grouping several feat/ / fix/devdev (PR)
feat/<name>one feature, a few commits (often one)dev or its sub/its parent (PR)
fix/<name>one fix, same as feat/dev, its sub/, or main for a production fixits parent (PR)
  • <name>: kebab-case, short, the subject of the work (feat/shared-memory-join, fix/argparser-crash, sub/network-v2). Delete the branch once merged.
  • When the repository is set up with gh (only on request), protect main so that it only accepts PRs:
    gh api -X PUT repos/<owner>/<repo>/branches/main/protection --input - <<'JSON'
    {"required_status_checks": null, "enforce_admins": true,
     "required_pull_request_reviews": {"required_approving_review_count": 1},
     "restrictions": null, "allow_force_pushes": false, "allow_deletions": false}
    JSON
    (required_approving_review_count / enforce_admins to adapt with the user; a free private repository can't use branch protection.)

Pull requests

Only for a project with several people, and only when the user asks (same rule as commits).

  • Target: the parent branch (feat/ & fix/ -> dev or their sub/, sub/ -> dev, dev -> main).
  • Title = commit subject format: feat(shm): add join for SharedMemory; a dev -> main PR is titled with the version: release: v2.14.0.
  • Description from templates/pull-request.md, in the user's writing style (like a release): a > [!NOTE] summary of 1-3 sentences, then the list of what was done CHANGELOG style (### Added / ### Changed / ### Fixed / ### Removed, one line per change, code in backticks, **[MAJOR]** for breaking changes), then ### Tests (what was run), optional > [!IMPORTANT] for what the reviewer must check. Build the list from git log --oneline <parent>..HEAD and the diff, never invent.
  • Classify it with gh: assignee, labels (create them if missing: feat, fix, docs, chore, test, breaking, release), reviewer, milestone when the project uses them:
    gh pr create --base dev --head feat/<name> --title "feat(<scope>): <message>" --body-file <body.md> \
        --assignee @me --label feat [--reviewer <login>] [--milestone <vX.Y.Z>]
    Ask the user who to assign / request a review from when it isn't them.
  • Never any AI attribution in the title, body, labels or comments.

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.