← ClaudeAtlas

vp-pr-comment-resolverlisted

Automate PR comment review, fix, and resolution workflow with atomic commits. Use when the user asks to "handle PR comments", "resolve PR review comments", "fix PR feedback", "process review comments", "address PR suggestions", "deal with review comments", or provides a GitHub PR URL with review comments. Also trigger when the user mentions unresolved PR threads, PR discussion comments at the bottom of the conversation, or wants to batch-process reviewer feedback. Boundary: not for writing PR reviews (use code-review) or PR checklists (use vp-checklist-runner).
VdustR/skills · ★ 1 · Code & Development · score 72
Install: claude install-skill VdustR/skills
# PR Comment Resolver Handle GitHub PR feedback by fetching review-thread comments and general PR discussion comments, verifying each request, applying focused fixes, committing by topic, replying with evidence, and resolving only the right review threads. ## Core Principles 1. **Critical thinking before action** - Reviewer comments may be wrong, outdated, ambiguous, or incompatible with repo rules. Verify against code, docs, and tests before acting. 2. **Commit by topic** - Group comments by logical change, not by comment count. Different concerns require separate commits. 3. **Preserve collaboration** - Ask the user before disagreeing with a reviewer, force pushing, changing scope, or resolving ambiguity. 4. **Reply to the correct target** - Review threads use GraphQL `addPullRequestReviewThreadReply`; PR discussion comments get a new PR comment that mentions the author and quotes the original excerpt. 5. **Resolve selectively** - Bot review threads are resolved after a decided outcome. Human review threads are left unresolved. PR discussion comments cannot be resolved. ## Quick Start Interactive mode: ```text User: Handle the comments on this PR: https://github.com/owner/repo/pull/123 ``` Default workflow: 1. Fetch unresolved review threads and PR discussion comments. 2. Exclude comments from the current viewer and prior resolver replies. 3. Classify each author as bot or human. 4. Verify whether each suggestion is valid, already handled, in