super-review:run

What problem does it solve?

It solves the problem of noisy, hallucinated, and out-of-scope AI PR reviews by producing a bounded set of actionable findings that are strictly tied to the PR diff and backed by quoted code evidence.

Core Features & Use Cases

• Evidence-quoted multi-agent review pipeline: runs parallel specialist reviewers, then enforces an evidence-confirmation gate before anything ships.
• False-positive and verification discipline: re-checks quoted findings against the actual code, then applies cross-reviewer collision checks and an Opus meta-verification pass.
• Diff-scoped, bounded reporting: keeps output strictly within the PR’s changed lines and caps the number of issues to reduce noise and reviewer overwhelm.
• Stack-aware sub-skill loading: auto-loads framework/security sub-skill catalogs based on detected technologies (for example React/Next.js/ORM/crypto/web headers/LLM security).

Quick Start

Run super-review in your Claude Code workflow by telling the assistant to review your pull request using the command "/super-review:run" or by pasting the GitHub PR URL and asking for a review.