What problem does it solve?
Iterating on code until it passes review normally requires manually alternating between a reviewer and a fixer, tracking findings across rounds, and knowing when to stop. This Skill automates that loop: it repeatedly runs a reviewer you name, applies fixes, guards scope, and escalates when fixes stop converging.
Core Features & Use Cases
- Autonomous review-fix loop: Launches a dynamic workflow that reviews, fixes, and re-reviews a target directory until a review reports zero critical/major findings.
- Cross-round findings ledger: Maintains an on-disk ledger so findings, rejections, and verdicts persist across rounds and across resumed runs after escalation.
- Scope guards and escalation: Mechanically enforces repo-relative scope globs, detects non-convergence (recurring findings, relocating bugs), and halts with structured violation reports instead of looping forever.
- Use Case: Point it at a module with
target=/repo/src/parser, reviewer plugin-dev:skill-reviewer, and scope src/parser/**; it iterates until the review is clean, then reports rounds used, remaining minor findings, and ledger/metrics paths.
Quick Start
Ask the AI to run an improvement loop over a directory, for example: improve the code in ./src/auth until review passes, using the pr-review-toolkit:review-pr skill as the reviewer with scope src/auth/**.
