What problem does it solve?
It helps teams catch bugs, security issues, missing tests, and risky assumptions in code changes before they ship, including when changes arrive via a local diff or a GitHub pull request.
Core Features & Use Cases
- GitHub PR paranoid review: Fetches PR metadata, unified diff, and per-file patch summaries, then produces findings with severity, category, and concrete fixes.
- Local diff review: Runs the same deep-review lenses for unstaged/staged diffs or the previous commit and reports findings in-chat.
- Line-level comment posting: For PRs, can post selected findings as RIGHT-side, line-anchored review comments using the PR head commit SHA.
- Six-lens quality checklist: Evaluates correctness, edge cases, security, test coverage, documentation/assumptions, and architectural concerns.
Quick Start
Ask the AI to run the code-review skill on owner/repo#123 to produce a paranoid architect review and recommend which findings to comment on.