Interactively generates commit messages, runs checks, and commits changes after confirmation.

Install

mkdir -p .claude/skills/commit-kabroxiko && curl -L -o skill.zip "https://agentskills.codes/api/skills/download/10274" && unzip -o skill.zip -d .claude/skills/commit-kabroxiko && rm skill.zip

Installs to .claude/skills/commit-kabroxiko

Activation

This is the description your AI agent reads to decide when to run this skill — the better it matches your request, the more reliably it fires.

Interactive `/commit` assistant command: propose commit message from git diff, stage all changes, and commit after confirmation.
128 charsno explicit “when” trigger
Intermediate

Key capabilities

  • Inspect repository status
  • Run automated checks
  • Generate commit messages
  • Stage changes
  • Execute commits

How it works

It inspects the repository, delegates checks to focused skills, generates a narrative commit message, and executes the commit upon confirmation.

Inputs & outputs

You give it
Staged or unstaged changes
You get back
Committed changes with generated message

When to use commit

  • Propose commit message from diff
  • Stage changes automatically
  • Run commit-time security checks

About this skill

When the user invokes the /commit command, the assistant will:

When the user invokes /commit, the assistant performs three distinct phases:

  1. Inspect the repository and working tree (sanity + changed files):

    • Run git status --porcelain --branch and git diff --name-only --diff-filter=AM HEAD to determine branch health and changed_set.
    • If a merge/rebase/conflict is in progress, abort and ask the user to resolve it.
  2. Run checks (delegated):

    • Rather than relying on a single helper script, /commit delegates checks to discrete, focused skills named below:

      • formatters — formatting and autofixes.
      • linters — linting and non-fixable errors.
      • dependencies — dependency vulnerability and outdated-package checks (NVD/OSV/GHSA-aware).
      • docker-scout — Dockerfile/image vulnerability scan using docker scout quickview.
      • secret-scan — secret scanning.
      • large-files — large file detection (>5MB).
      • eslint-provision — optional ESLint provisioning when JS/TS files changed.
    • Each check skill describes its invocation, exit codes, and when the assistant must pause for confirmation.

      • When the /commit orchestrator invokes these checks it MUST call the checks skill with scope=changed so sub-skills operate only on the computed changed_set.

      • The /commit orchestrator accepts an optional --skip-checks (or skip_checks) flag. When provided, the assistant will skip invoking the checks skill entirely and proceed directly to commit-message generation and staging. This is intended for advanced workflows where the caller intentionally bypasses automated checks; use with caution.

        NOTE: checks here refers to an assistant/skill invocation, not an OS shell program. The assistant should invoke the repository's checks skill (for example, via the assistant's skill/capability API), and must not attempt to run a literal checks command in the user's shell. Example skill-like invocations (conceptual):

        • checks --scope=changed (invoke the checks skill to run all checks against changed files)
        • Or skip checks when invoking /commit by passing --skip-checks, e.g. /commit --skip-checks.
        • checks --check=linters --scope=changed (invoke the checks skill to run only linters against changed files)

        If the checks skill is not available, the assistant should skip running checks and continue to the commit-message generation and staging steps.

  3. Commit message generation and commit execution:

    • The assistant synthesizes a multiline commit message based on the actual content changes (the diff), not simply a list of filenames, and presents it for user confirmation/edit. The assistant MUST inspect the diffs for the computed changed_set (for example via git diff --staged or git diff HEAD -- <paths>) and summarize what was changed and why at a code/content level. The required format is:

      • A concise one-line header (summary) — preferably <= 50 characters.
      • A blank line.
      • An unstructured narrative paragraph (one or more sentences) describing the change in freeform prose. The narrative should explain the reason for the change, what was modified (behaviour, API, tests, docs), and any developer- or user-visible impact. When relevant, mention added or updated tests and migration steps.
    • Guidance for message content:

      • Prefer summarizing the content of diffs: code paths modified, bugs fixed, logic added or removed, and tests added. Do not default to enumerating filenames; only include filenames when they provide helpful context (e.g., "update config loader in settings.go").
      • Highlight user-facing changes, breaking changes, or required follow-ups (e.g., run go mod tidy, database migrations, or npm install).
      • Keep the header atomic and the body explanatory and actionable for other developers.

      Example:

      Fix crash on startup when config missing

      The app attempted to read an unset env var and panicked; add a safe fallback and unit tests to cover the code path. No user-facing impact; developers should run make test after pulling.

    • For the final staging/commit step the assistant will: - Present the proposed commit message and a summary of the changed files for user confirmation; by default, when the user approves, stage all changes. - If the user approves, execute the necessary git commands directly in the working tree, for example:

      git add --all
      # Use a heredoc to provide a multiline commit message safely
      git commit --file=- <<'COMMIT_MSG'
      <One-line header>
      
      <Unstructured narrative describing why the change was made, what changed, and any impact>
      COMMIT_MSG
      
      • If the user prefers to run commands locally, provide the exact git commands to run and do not execute them automatically.

Rules & constraints:

  • Never push to remotes automatically.
  • Each check is a separate skill and must be implemented or provided by the repository owner if the repo requires deterministic automation.
  • The assistant must not conflate multiple checks into a single script without explicit repository policy that mandates it.

See the sibling checks-* skills for details on each check's behavior and exit codes. - checks/formatters — formatting and autofixes.

When not to use it

  • Pushing to remotes automatically
  • Committing during active merge/rebase

Prerequisites

checks skill

Limitations

  • Does not push to remotes automatically

How it compares

It automates the entire commit preparation and validation process instead of manual staging and message writing.

Compared to similar skills

commit side by side with the closest alternatives in the catalog.

SkillInstallsUpdatedSafetyDifficulty
commit (this skill)02moNo flagsIntermediate
resolve-conflicts818moReviewIntermediate
git-advanced-workflows112moReviewAdvanced
nx-workspace46moReviewBeginner

Try saying

Example prompts that trigger this skill in your AI assistant.

Search skills

Search the agent skills registry