What problem does it solve? Code written by an AI agent is often committed without independent verification, letting security flaws, logic errors, and regressions slip through. This Skill enforces a pre-commit verification pipeline so no agent verifies its own work. ## Core Features & Use Cases - Static Security Scanning: Greps added diff lines for hardcoded secrets, shell injection, eval/exec, unsafe pickle deserialization, and SQL injection patterns. - Baseline-Aware Quality Gates: Runs pytest, npm test, cargo test, or go test plus linters (ruff, mypy, eslint, clippy, go vet), comparing against a pre-change baseline so only new failures block the commit. - Independent Reviewer Subagent: Dispatches a fresh-context reviewer via delegate_task that returns a fail-closed JSON verdict on security concerns and logic errors. - Auto-Fix Loop: Spawns a separate fix agent for up to two fix-and-reverify cycles before escalating to the user. - Use Case: After implementing a feature across several files, trigger this Skill before running git commit so the diff is scanned, tested, independently reviewed, and committed with a [verified] prefix only if all gates pass. ## Quick Start Verify my staged changes with the pre-commit review pipeline before committing them.
