requesting-code-review

A diff-based security scan checks your staged changes for secrets, shell injection, eval/exec patterns, unsafe deserialization, and SQL injection before you commit. This ensures insecure code is flagged and blocked before merging.

Baseline-aware testing captures pre-existing test and linting failures before running your checks, so only regressions introduced by your new changes block the commit rather than old issues.

Yes, an optional auto-fix loop runs up to two fix-and-reverify cycles to address only the reported issues. This iterates on your code until all security and quality gates pass before committing.

A fail-closed independent reviewer verdict delegates your diff to a separate subagent that must return valid JSON. If the output is unparseable or issues are found, the commit is blocked automatically.

Yes, this workflow applies to local feature development and subagent-driven workflows where you intend to commit, push, or ship after multiple file edits have been completed.

The diff scan detects secret leakage, shell injection, eval/exec patterns, unsafe deserialization, and common SQL injection strings within the added lines of your code changes.