By name: code-review description: Run a thorough self-review pass on the most recent change. allowed-tools: [read, bash] permissions: bash: ["git diff*", "git log*", "git show*", "git status"]
Code review
When the user asks for a code review (and you have not already done one in this turn), follow this routine.
1. Establish what changed
Use bash to run git status and git diff (or git diff --staged
if there are no unstaged changes). Skim the patch end to end before
analysing any single hunk.
2. For each modified file
Read the file in full with the read tool — never review only the
hunk; you need surrounding context to evaluate the change properly.
Then look for:
- Correctness: bugs, off-by-one errors, wrong sign, missing nil checks, swapped arguments, race conditions.
- Error handling: every external call (file IO, network, parsing, syscalls) — does it propagate or swallow? Are errors wrapped with enough context?
- Tests: do the new code paths have tests? Are existing tests still passing what they claim?
- Surface area: are exports necessary, or could the change stay internal? Public APIs deserve more scrutiny than internals.
- Style consistency: does the change match neighbouring code?
3. Report
Produce a concise written review with this shape:
- Verdict (one line): ship-as-is / minor changes / needs work / blocked.
- Required changes (numbered, if any).
- Suggestions (bullets, optional).
- Praise (one or two lines if anything stood out — keeps the feedback humane).
Don't restate every line of the diff. Don't speculate about future features. Stay grounded in what the patch does.
4. Stop
Do not auto-apply fixes. The user will decide what to act on.