What problem does it solve? Reviewing a busy GitHub pull request means manually reading every issue-level, review-level, and inline comment, deciding which ones require code changes, which deserve a polite reply, and which are bot noise. This Skill automates that triage so nothing actionable is missed and no time is wasted on outdated or duplicate feedback. ## Core Features & Use Cases - Full comment harvesting: Pulls all PR comments via the GitHub CLI (gh), including issue comments, reviews, and inline file comments, then de-duplicates them into one list. - Structured classification: Sorts each comment into bot vs human, then into actionable buckets such as correctness, security, scope creep, question, nit, discussion, outdated, or praise. - Triage report with drafts: Produces a prioritized report containing an implementation plan for valid comments and pre-written, professional English responses for comments to decline, without ever posting to GitHub until explicitly approved. - Use Case: A reviewer leaves 25 comments on your PR mixing lint bot warnings, a security concern, and style nits. Run this Skill with the PR URL to get a sorted report: 2 critical fixes with a commit plan, 3 drafted replies, and the rest safely ignored. ## Quick Start Pass a GitHub PR URL and ask the agent to triage all review comments and draft responses for the ones that should not be implemented.