/commit

← Back to index

Summary

Stop writing commit messages by hand. Reads the staged diff, produces a Conventional Commits message with a why-first body, and either commits immediately or hands you the text to paste — your call.

When to use

You've finished a change and need a commit message that explains the why, not just the what. Use /commit to generate and commit in one step, or /commit message to get the text without touching the repo.

Trigger phrases (from SKILL.md):

Two modes: bare /commit generates and commits; /commit message drafts the text only — never mutates the repo.

Invocation

Commit mode — generate and commit:

/commit

Draft mode — message text only, no git mutation:

/commit message

The skill reads the staged diff (or unstaged if nothing is staged). Message shape: <type>(<scope>?): <imperative description> subject line + one or more ## H2 body sections, each with a context paragraph then bullets. Types: feat, fix, docs, refactor, chore, test, perf.

Example 1 — Commit mode (illustrative example)

Input:

/commit

Output / Result:

Commit mode: I will commit after generating the message.

```
docs(readme): document 8 previously undocumented skills

## README coverage

Eight skills existed in `skills/` with no entry in the "What it
gives you" section. Readers had no way to discover them without
opening raw SKILL.md files directly.

Added a pain-first bullet for each skill, placed in the section
that best matches its purpose:

- "Ship a feature, start to finish": `/quick-plan`, `/commit`,
  `/wrap-up`
- "Get a second opinion": `/clean-code-review`, `/options`
- "Wrangle docs and skills": `/doc-voice`, `/md-reviewer`,
  `/plain-language`
```

6c5d1ed docs(readme): document 8 previously undocumented skills
 README.md | 8 ++++++++

Example 2 — Draft mode (illustrative example)

Input:

/commit message

Output / Result:

Draft mode: I will generate the message only — no commit.

```
feat(api): add POST /users/bulk endpoint for batch creation

## Batch user creation

Creating users one at a time via POST /users was too slow for
import flows — a 500-row CSV took ~25 s of sequential requests.
The new endpoint accepts up to 1 000 user objects in one call.

- New route: `POST /users/bulk` in `src/routes/users.py`
- Validates each row individually; partial failures return per-row
  errors without rolling back successful rows
- Added integration tests: `tests/test_users_bulk.py` (12 cases)
```

Paste into `git commit -F-` or `git commit -e` to review.

Related tools