requesting-code-review

To run a pre-commit code review, you ask an AI assistant with the Skill to verify your staged git diff. It executes security scanning, linting, and baseline-aware tests, returning a fail-closed JSON verdict before you commit.

Automatically checking staged git diff for security and regressions involves running static security scans on added lines and baseline-aware tests. An independent reviewer then provides a fail-closed JSON verdict to block commits if new failures appear.

You can automatically fix linting and security issues using an optional auto-fix loop. It attempts up to two fix-and-reverify cycles to resolve reported problems without refactoring or adding features before you push.

The pre-commit verification pipeline supports skipping for documentation or config-only changes. This prevents unnecessary baseline-aware test runs and static security scans when no executable code logic is modified.

An independent reviewer subagent is needed for pre-commit code verification to ensure an unbiased, fail-closed JSON verdict based strictly on the diff and scan results, avoiding shared context contamination from the implementation phase.

The auto-fix loop in pre-commit code verification is limited to two fix-and-reverify cycles. It only attempts to resolve reported issues without refactoring code or adding new features, stopping if issues persist.