git-commit · v1.0.0 · 2026-07-19 · sha256 134bcb8316b4413c

git-commit v1.0.0A

Immutable. This exact content is served forever at /api/v1/blob/134bcb8316b4413c.

---
name: git-commit
description: 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).
metadata:
  author: mgiovani
  version: 1.0.0
disable-model-invocation: false
argument-hint: ''
allowed-tools:
- Bash(git *)
- Read
- Edit
- Write
hooks:
  PreToolUse:
  - matcher: Bash(git commit*)
    hooks:
    - type: command
      command: ${CLAUDE_PLUGIN_ROOT}/skills/git-commit/scripts/pre-commit-lint.sh
      timeout: 60
---

# 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:
1. Read `git status` and `git diff --staged` (or `git diff` if nothing is staged yet) before writing anything.
2. Check which files/modules changed before setting scope.
3. Look for removed exports, changed signatures, deleted functions to catch breaking changes.
4. 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

1. Run `git status` and `git diff --staged` to see what actually changed.
2. 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.
3. 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
4. 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.