What problem does it solve?
Many large language models have training-specific blind spots that can miss security vulnerabilities; this Skill provides a protocol to obtain an independent, uncontaminated review from a second model so you can catch findings the primary model missed and reduce false confidence on critical changes.
Core Features & Use Cases
- Independent Review: Route the change diff and relevant context to a separate LLM session without exposing the primary model's findings to ensure an unanchored analysis.
- Comparison & Triage: Produce a comparison matrix of findings, highlight agreements and disagreements, and prioritize disagreements for human investigation.
- Use Case: Ideal for authentication changes, payment flows, PII handling, cryptographic code, or any change flagged CRITICAL by earlier automated layers.
Quick Start
Route the change diff and supporting context to a fresh LLM session and ask it to perform a security review and list findings without revealing prior model outputs.