What it does
Claude Code ships with bundled skills out of the box — not fixed logic but prompt-based skills: they hand the model detailed instructions and let it orchestrate the work with its own tools. Available in every session via /name. The key ones for development:
/code-review— reviews the current diff for bugs and simplifications; effort levels (low/medium → targeted, high → broader), plus the flags--comment(inline comments in the PR) and--fix(apply the changes)./security-review— a security audit of your changes./review— pull request review;/init— initialisation. Those three (/init,/review,/security-review) are also available programmatically through the Skill tool./run,/verify— launch the app and confirm the change works in the live app, not just in tests.
Adding your own /commit skill is easy: the docs show a skill with disable-model-invocation: true and allowed-tools: Bash(git add *) Bash(git commit *).
How to set it up
Nothing to install — the bundled skills are already in Claude Code (you need a recent version; /run and /verify require v2.1.145+). Put your own wrapper skill (say /commit or /review-pr) in ~/.claude/skills/<name>/SKILL.md.
When to use it
- Before a commit or PR — a quick
/code-reviewand/security-reviewpass over the diff. - On risky or multi-file changes — a mandatory cross-check.
- To confirm a fix against the live app —
/verify.
Example
After making changes, type /code-review --fix — the model reads the diff, finds the bugs and applies the fixes to your working tree itself; then run /security-review for the audit.