commit

An Agent Skill that teaches an agent to turn the changes in any git working tree into clean Conventional Commits: read the diff, run the repository's own check, stage by path, split unrelated changes into separate commits, write the message with the markers the release tooling reads, and verify what landed. It works in any repository and any stack, the Next.js apps of the other Timerise Skills included.

The history is the release input. The release skill bumpv decides the next version from the commits alone, so a commit that hides a breaking change, sweeps in a secret or carries a tool's trailer ships that mistake in the next release. Stage by path, write the markers, and check what landed.

This skill was written by the engineer who maintains the Timerise Skills and releases every one of them from the history it writes. The earlier implementation was a single-file version of the skill. What it carries holds by the commands it runs: every file staged by explicit path and listed with git diff --staged --stat before the commit; every breaking change marked with both ! and a BREAKING CHANGE: footer; every commit's message read back with git log -1 --format=%B, so no attribution trailer stays; and no amend after a failed hook or on a pushed commit. references/provenance.md has the record.

Install

One command, via the skills.sh CLI, which installs the skill into every skills-compatible agent it detects, including Claude Code, Codex CLI and Gemini CLI:

npx skills add timerise-ai/commit

Name the agents instead with -a, for example npx skills add timerise-ai/commit -a claude-code -a codex.

Manual install

Nothing here is Claude-specific: the skill is a plain Agent Skills folder, SKILL.md plus markdown references with no file that calls a model, so cloning it into an agent's skills directory is all an install is. For Claude Code:

git clone https://github.com/timerise-ai/commit.git ~/.claude/skills/commit

To scope it to a single project instead, clone it into that project's .claude/skills/ directory. For another agent, clone into that agent's skills directory, or symlink the Claude Code copy so one git pull updates every agent:

mkdir -p ~/.agents/skills
ln -s ~/.claude/skills/commit ~/.agents/skills/commit

Update the skill with git pull in its directory. The current release is 0.1.1. See CHANGELOG.md. The skills index lists the other Timerise Skills and how to install them all at once.

Activation

The skill activates automatically when the user asks to commit, save progress, record WIP, split the working tree into commits, or wrap up a finished task. Invoke it explicitly with /commit in Claude Code, $commit in Codex CLI, or from /skills in Gemini CLI.

Each host matches a task against the description its own way, so invoke the skill explicitly on a first run rather than assuming it fired. Only SKILL.md is read up front; the references/ files load on demand.

What's inside

File	Contents
SKILL.md	Entry point: the flow diagram, the seam, critical facts, hard rules, quick start and reference directory
references/staging.md	Reviewing the tree, untracked files, what is never staged, grouping, staging by path, a file with two changes
references/validation.md	Finding the repository's check per project type, when to skip it, and reading its result
references/messages.md	Types and how bumpv reads them, scope, subject, body, footers, breaking changes, examples
references/committing.md	The heredoc commit, verifying what landed, hooks, when amending is allowed, repository states, the report
references/adaptation.md	The seam contract: the host's conventions and where they win, the contract with bumpv, the order of work, the non-negotiables
references/provenance.md	The engineering ledger: what the audit changed, what was kept deliberately, and what was added here
README.md	This file
CHANGELOG.md	One section per release, newest first
CLAUDE.md	The editing conventions, for an agent editing this repository
LICENSE	MIT
evals/	The prompts an operator types after installing (prompts.md) and one file per agent eval: the skill installed into an empty Next.js app, one prompt that asks for a change and its commit, no help, then type-checked and built
.github/workflows/agent-eval.yml	The caller of the index's reusable eval workflow, run on every published release and on a maintainer's dispatch

The skill builds no module, so its evals score the agent's fidelity to the hard rules, not the app: the checks only confirm the app still type-checks and builds, and the notes on each run carry the score. The seam with the host repository is its own check command, its scope vocabulary and its documented commit conventions, read and never rewritten. A stricter host convention is followed; one that would break a non-negotiable is reported to the user.

The six non-negotiables

These are never optional. Each is stated as a hard rule in SKILL.md, in the same order:

1. Never put AI attribution in a commit. The history is read by people and by release tooling, and a tool's trailer belongs to neither; git log -1 --format=%B after every commit verifies it, and an amend removes one that a harness or hook added.
2. Stage by path, never with git add ., -A or git commit -a. Each sweeps in whatever is lying in the tree; secrets, build output and large binaries are flagged and never staged, and git diff --staged --stat lists the commit before it is made.
3. Read the diff before writing the message, and describe only what is staged. The message is what the release notes are written from, so it is checked against git diff --staged, not against the task.
4. Mark every breaking change twice. bumpv picks a MAJOR from the markers alone; the ! shows in git log --oneline and the BREAKING CHANGE: footer carries the migration path into the changelog.
5. Never bypass or misdirect a hook. A hook is the repository's rule, so no --no-verify; a failed hook leaves no commit, so an amend after it rewrites the previous commit; and a pushed commit is never amended.
6. Never push, tag, bump a version or edit the changelog. The release belongs to bumpv, which writes chore(release):; pushing is the user's call.

Everything else is the host repository's: its check, its scopes, its commit rules, its hooks.

Not this

Not this	Use instead
Bumping the version, writing the changelog, tagging a release	bumpv
Pushing, or opening and describing a pull request	the user, after this skill reports
Rewriting published history: rebase, squash, force push	a deliberate, separately confirmed step
Reviewing the change for bugs before it is committed	a code review skill
Hardening a skill from its agent evals	skill-eval-loop

Contributing

Issues and pull requests are welcome here. Pure markdown, with no build step and nothing that executes; the commands in references/ run in the repository being committed in, not this one. Claims in this skill are meant to be verifiable: if you change a factual claim about git, a hook or the commit format, say how you verified it, whether against git's own documentation (githooks(5), git help commit), the Conventional Commits specification, bumpv's decision criteria, or a reproduction in a scratch repository.

Adding, removing or renaming a file in references/ means updating the quick start and the reference directory table in SKILL.md, the file table above, and any relative cross-links. references/provenance.md is the ledger that must stay truthful; add an entry for anything you change. Commits follow Conventional Commits and releases follow STANDARD.md in the index; CLAUDE.md carries the full editing conventions.

Part of the Timerise Skills

This is one of the Timerise Skills: modules for Next.js App Router apps written by our own senior engineers from the modules they have shipped, not synthetic, each published as its own repository and indexed there. They share one layout, so an agent that has read one knows how to read the next: a SKILL.md entry point, references/ loaded on demand, and a seam contract carrying the module's non-negotiables. This one is for any repository: it writes the history the skills are released from.

Author

Built and maintained by Timerise.

License

MIT. See LICENSE.