What problem does it solve?
Reviewers — especially AI models — produce confident, specific findings that are wrong at a meaningful rate, and blindly applying their suggested fixes can break correct work. This Skill turns incoming review comments into evidence-backed decisions instead of unverified edits.
Core Features & Use Cases
- Finding triage and verification: Classify each finding by type and severity, then confirm, refute, or partially confirm it against the actual source with line-level evidence.
- Mechanical checks over opinion: Run computable checks — grep symbols, resolve cross-references, execute consuming code — before trusting reviewer confidence.
- Poisoned-fix detection: Evaluate the proposed fix, not just the finding, to avoid introducing real defects while addressing cosmetic ones.
- Use Case: You receive a 40-item AI referee report on a research paper. Use this Skill to verify each claim against the manuscript, batch the confirmed fixes, refute misreads with cited text, and add signposts where reviewers stumbled.
Quick Start
Adjudicate the attached referee report against my paper, verifying each finding before making any changes.
