review

Run the review workflow on your feature branch. It fetches the latest base branch, generates the full diff, and applies a two-pass checklist covering SQL safety, race conditions, LLM trust boundaries, and informational issues like dead code and test gaps.

It catches structural issues tests do not cover: SQL string interpolation, TOCTOU race conditions, unvalidated LLM output written to databases, enum values not handled by all consumers, and conditional side effects where one branch skips an action.

Yes. The workflow detects the platform from the git remote URL and uses glab CLI for GitLab or gh CLI for GitHub. If neither CLI is available, it falls back to git-native commands to determine the base branch.

It fetches Greptile line-level and top-level PR comments via the GitHub API, classifies each as valid, already fixed, false positive, or suppressed based on project history, then posts evidence-backed replies using tiered templates.

Design review runs only when the diff touches frontend files, detected via the pstack-diff-scope tool. It checks for AI slop patterns, typography issues, spacing problems, and missing interaction states, calibrated against DESIGN.md if present.

Diff-based review cannot verify runtime behavior, so it flags issues for human judgment on security, race conditions, and design decisions rather than auto-fixing them. Enum completeness checks require reading files outside the diff, which adds review time.