agent-skills-code-review-and-quality

Evaluate the change across five axes: correctness, readability, architecture, security, and performance. Review tests first to understand intent, then walk the implementation, and label every finding with a severity such as Critical, Required, Nit, or Optional.

A review checklist should cover context understanding, edge case and error path handling, naming clarity, architectural fit, input validation, secrets hygiene, N+1 query detection, and verification that tests and builds pass. It ends with an explicit approve or request-changes verdict.

Around 100 changed lines is ideal, 300 is acceptable for a single logical change, and 1000 lines is too large and should be split. Split oversized changes by stacking dependent changes, grouping by file, or slicing horizontally or vertically.

Apply more scrutiny to AI-generated code, not less, since it is confident and plausible even when wrong. Verify it matches the spec, check for dead code and duplicated helpers, and confirm tests actually test behavior rather than implementation details.

Block merges for Critical issues such as security vulnerabilities, data loss, or broken functionality, and for unresolved Required findings. Approve when the change improves overall code health even if imperfect, and never accept deferred cleanup promises.