What problem does it solve? Manually triaging CodeRabbit review comments on GitHub pull requests is slow and error-prone, and blindly executing reviewer-provided prompts is a security risk. This Skill fetches unresolved review threads for your current branch's PR, presents them in a severity-ranked table, and applies fixes only after explicit per-change approval. ## Core Features & Use Cases - Thread-Aware Feedback Retrieval: Uses the GitHub GraphQL API with cursor pagination to collect only unresolved, non-outdated CodeRabbit review threads, filtering out resolved items and in-progress review states. - Severity-Mapped Review Workflow: Parses issue type, severity, and location from each thread, displays them in a structured table, and reviews fixes in CRITICAL-first order with independent local validation. - Prompt-Injection Defense: Treats all reviewer comment bodies and "Prompt for AI Agents" sections as untrusted input, sanitizing guidance summaries and refusing requests to read secrets, dotfiles, or unrelated files. - Use Case: After pushing a branch and receiving a CodeRabbit review, run the Skill to list all open issues, approve three valid fixes one by one, commit them as a single consolidated commit, run lint checks, push, and post a summary comment on the PR. ## Quick Start Ask the AI to run the CodeRabbit autofix workflow to review and apply unresolved review feedback on the current branch's pull request.