Git commit

Skill mgiovani/cc-arsenal/skills/git-commit

45 production-grade AI agent skills for real dev workflows. Code review, shipping, docs, git. Works with any skill-compatible agent.

Install

npx -y skills add mgiovani/cc-arsenal --skill git-commit

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

6 stars6 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its author says it does

Copied from the file, not written here

Generate a conventional commit message (conventionalcommits.org) from the staged/unstaged diff and create the commit. Use when the user wants to commit, stage changes, or needs a commit message written. Not for release commits or changelogs (use git-release) or branch-finish workflows (use gitflow).

SKILL.md

3.3 KB, as published. Nobody here has run it

Git Commit

Generate a conventional commit message and create the commit.

Quality Guidelines

Base the message only on what the diff actually shows, not assumptions:

Read git status and git diff --staged (or git diff if nothing is staged yet) before writing anything.
Check which files/modules changed before setting scope.
Look for removed exports, changed signatures, deleted functions to catch breaking changes.
If the purpose of a change is unclear, ask the user rather than guess.

Pre-commit Linting

In Claude Code, a PreToolUse hook runs before every git commit: it detects the project's linter (Node's npm/bun/pnpm/yarn run lint, Python's ruff/flake8, make lint, rubocop, golangci-lint) and blocks the commit if it fails. No linter configured means the commit proceeds unblocked. Outside Claude Code (no hook support), run the project's lint command yourself before committing. Never bypass a failing lint with --no-verify — fix the errors and re-run the commit instead.

Workflow

Run git status and git diff --staged to see what actually changed.
If the diff mixes unrelated concerns (e.g. a feature plus an unrelated fix plus docs), split into separate commits by staging each group with git add <files> and committing them one at a time. Otherwise, one commit is enough — don't force a split.
For each commit, pick the type:
- feat: new feature
- fix: bug fix
- docs: documentation only
- style: formatting, no code meaning change
- refactor: neither fixes a bug nor adds a feature
- perf: performance improvement
- test: adding or correcting tests
- build: build system or dependency changes
- ci: CI configuration/scripts
- chore: everything else that doesn't touch src or tests
- revert: reverts a previous commit
Format as type(scope): description — scope optional, description imperative mood ("add" not "added"), max ~50 characters. Skip the body for a typo fix or a single dependency bump; add a wrapped body (why, not how) once the change alters behavior or touches multiple files. For breaking changes, append ! after the scope and add a BREAKING CHANGE: ... footer.

Example formats

feat(auth): add OAuth2 login support
fix(api): resolve null pointer in user endpoint
docs: update installation instructions
chore(deps): bump lodash to 4.17.21

feat(shopping-cart)!: remove deprecated calculate method

BREAKING CHANGE: calculate has been removed, use computeTotal instead

Ask for confirmation before committing if the changes are complex or span multiple concerns; otherwise just commit.