audit-review

Verify PR review findings, sweep defect classes, and resolve review threads on GitHub.

Updated Sep 23, 2026
One-click install
npx skills add https://github.com/drgaciw/rlm-base-dev-fork --skill audit-review-drgaciw
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: audit-review
Source: https://github.com/drgaciw/rlm-base-dev-fork/tree/main/.cursor/skills/audit-review
Command: npx skills add https://github.com/drgaciw/rlm-base-dev-fork --skill audit-review-drgaciw

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve? Automated PR reviewers like Codex and Copilot post findings that are sometimes wrong, duplicated, or only partially scoped. This Skill provides a disciplined playbook for verifying each finding against the actual code, sweeping the entire defect class across the change, and closing every review thread so a branch survives a strict pre-merge audit. ## Core Features & Use Cases - Finding triage and verification: Classify each bot comment as real, partial, or false-positive with evidence, and group duplicate findings across reviewers. - Completeness sweeps: Fix every instance of a valid defect class (e.g., missing WITH USER_MODE in SOQL) across the whole feature, not just the cited line. - Thread resolution protocol: Reply in-thread with the commit SHA, react to valid comments, resolve threads via GraphQL, and confirm zero unresolved threads across all pages. - Pre-merge main comparison audit: Detect branch-side reverts of newer main content and foreign commits inherited from composed integration branches before merging. - Use Case: Before promoting a release branch to main, run the deep audit: triage 15 bot comments into 5 classes, sweep each class repo-wide, push one cohesive commit, and close the round at zero unresolved threads. ## Quick Start Ask the agent to process the automated review comments on your pull request using the audit-review playbook and resolve every thread.

Frequently Asked Questions about audit-review

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I handle automated PR review comments from Copilot or Codex?▼

Verify each finding against the actual source code and classify it as real, partial, or false-positive. Fix valid findings as a whole class across the feature, reply in-thread with the commit SHA, react to valid comments, and resolve the thread via GraphQL.

How to resolve GitHub review threads with the gh CLI?▼

GitHub's REST API cannot resolve review threads, so use the GraphQL resolveReviewThread mutation via gh api graphql. List threads with the reviewThreads query under repository and paginate with endCursor until hasNextPage is false.

Should I trust automated code review findings at face value?▼

No. Bots are sometimes wrong, so confirm each claim by reading the cited file and lines before changing code. Refute false positives with evidence in a thread reply rather than patching correct code.

How do I detect a branch that reverts newer main changes before merging?▼

Diff the branch against main ranked by deletions, since removal-dominated files signal a revert. Cross-check files recently developed on main that the branch also changed, and confirm whether a specific main fix was dropped before resolving.

What are the limitations of an offline code review by an external agent?▼

Offline reviews cannot catch data-shaped defects, reproduce reported test figures, or establish platform engine behavior from source. Run the code live after a review round, not just the test suite, and flag unverifiable premises explicitly.