What problem does it solve? Pull request feedback arrives scattered across SonarQube scans, CodeRabbit reviews, and human reviewer threads, and addressing it manually means re-checking each source, tracking what was already fixed, and risking pushes to the wrong branch or a merged PR. This Skill gathers all feedback sources, prioritizes findings, applies fixes, and replies to and resolves review threads with verified dispositions. ## Core Features & Use Cases - Multi-source feedback aggregation: Fetches findings from SonarQube, CodeRabbit, and human reviewers in parallel, skipping any source that is unavailable. - Safety-gated pre-flight checks: Verifies the local checkout matches the PR head commit, branch, upstream, and remote push URLs before editing or pushing, and refuses fork PRs, merged PRs, and dirty working trees. - Iteration and round-budget awareness: Detects previously handled comments via disposition markers, enforces a configurable agent-to-agent round budget, and collapses repeated same-class findings into one structural fix. - Post-merge fix-forward flow: When feedback lands after a merge, branches off the base branch, fixes there, and opens a follow-up PR instead of resurrecting a dead head branch. - Use Case: After CI posts SonarQube issues and CodeRabbit leaves inline comments on PR #286, run the Skill to fix each finding in priority order, commit, push, and reply to every thread with its disposition. ## Quick Start Ask the AI to run ar-taskflow-fix-comments on the current pull request to gather all reviewer feedback, apply the fixes, and resolve the threads.
