What problem does it solve? Code review often misses real breakage hidden outside the diff, produces noisy low-signal comments, or skips deploy-risk checks entirely. This Skill reviews a diff, PR, branch, or commit range with full context — changed files, callers, contracts, and deploy surface — and returns evidence-backed findings ranked P0–P3 plus a ship gate. ## Core Features & Use Cases - Severity-ranked findings: Every finding carries a file:line, evidence, a concrete fix path, and a confidence level, ordered P0 (data loss, security exposure) through P3 (cleanup). - Deploy Safety Gate and Commit Dirt Score: Automatically grades breaking changes, rollback safety, blast radius, env readiness, and scans the diff for secrets, debug artifacts, and conflict markers via a bundled shell script. - Stack-specific trap catalogs: Reference files cover TypeScript, React, Next.js, accessibility, SEO, database, Swift/iOS, OWASP 2025 baselines, and a twelve-smell maintainability baseline, loaded only when the diff touches that stack. - Use Case: Before merging a PR that touches auth and a database migration, run a --deep review to trace cross-file call chains, check OWASP baselines, and get a ship / ship-with-caveats / hold verdict with named blocking findings. ## Quick Start Ask the agent to use suede-code-review to review the current branch diff against main at standard depth and report findings with the deploy safety gate.
