szymonkaliski/git-commit

Commit all changes in the repo with proper formatting, linting, and testing.

1. Gather Context

Check what tooling exists:

!ls -la package.json Makefile Cargo.toml pyproject.toml go.mod 2>/dev/null || true

!cat package.json 2>/dev/null | head -50 || true

Study the repo's commit message style:

!git log --oneline -20 2>/dev/null || true

Note the pattern: conventional commits? imperative mood? lowercase? prefix? scope? Look at casing, punctuation, length, and structure. Your commit message MUST follow this pattern.

2. Current Changes

!git status

!git diff

!git diff --cached

3. Execute

Based on what you found:

1. Format: Run formatter if available (prettier, black, gofmt, cargo fmt, nixfmt, etc.)
2. Lint: Run linter if available (eslint, tsc --noEmit, clippy, etc.)
3. Test: Run tests related to changed files if test runner exists
4. Plan commits: Look at the git log style. If the repo favors small, focused commits (one logical change each), split the staged/unstaged changes into multiple commits — group related files together by logical change. If the repo uses larger commits, a single commit is fine.
5. Stage & commit: For each logical group, git add the relevant files by path (never git add -A — avoid staging .env, credentials, or large binaries) and commit with a message matching the repo's style. Go from most independent change to most dependent.

If $ARGUMENTS is provided, use it as the commit message for a single commit of all changes. Otherwise, generate messages matching the repo's commit style, splitting into multiple commits when it matches the repo's pattern.

Fix any formatting/lint issues automatically. If tests fail, report and stop.