What problem does it solve?
This skill prevents risky or incorrect code from getting merged by analyzing a pull request diff for SQL safety, LLM output trust boundary violations, conditional side effects, race conditions, and other structural issues that commonly slip past. It helps you catch problems early so you avoid broken behavior, security exposures, and hard-to-debug incidents in production.
Core Features & Use Cases
- Pre-landing diff analysis: Inspects the changes against the base branch with a structured checklist and focuses on problems most likely to matter in real deployments.
- SQL and data safety checks: Flags unsafe query construction, non-atomic read-check-write patterns, validation bypasses, and missing eager-loading that causes N+1 behavior.
- LLM trust boundary checks: Reviews where model-generated values, tool outputs, or URLs are accepted without lightweight validation or allowlist controls, reducing prompt injection and SSRF risk.
- Race condition and concurrency review: Detects unsafe status transitions and duplicate-creation patterns that require atomic updates or uniqueness constraints.
- Shell injection and enum/value completeness checks: Looks for dangerous command invocation patterns and ensures new enum-like values are handled across all consumers.
Quick Start
Use this skill when your request says to review this PR or check my diff, focusing on the current changes you plan to land.
