review

Run the review workflow on your feature branch; it fetches the base branch, generates the full diff, and applies a checklist covering SQL injection, XSS, race conditions, and unvalidated LLM output. Findings are cited with file:line references and suggested fixes.

It catches structural issues like TOCTOU race conditions, missing eager loading, enum values not handled in all consumers, conditional side effects, magic numbers, and CI/CD pipeline misconfigurations. These are categorized into critical and informational severity passes.

Yes, mechanical fixes like dead code removal, magic number extraction, and stale comment updates are auto-applied. Riskier changes involving security, race conditions, or user-visible behavior are batched into a question for your judgment.

Yes, it reads TODOS.md, PR descriptions, commit messages, and plan files to extract actionable items, then cross-references them against the diff. Items are classified as DONE, PARTIAL, NOT DONE, or CHANGED, and out-of-scope changes are flagged as drift.

Yes, it fetches Greptile line-level and top-level comments via the GitHub API, classifies them as valid, already fixed, false positive, or suppressed, and posts evidence-based replies. This step is skipped silently if no PR or comments exist.

The design review runs only when the diff touches frontend files, detected via a scope script. It checks for AI-slop patterns, typography issues, missing focus states, and DESIGN.md violations, and skips silently for backend-only changes.