What problem does it solve? Closing out a pull request requires an unbiased code review plus fixes for every piece of feedback, but doing both in one session lets the author's own context bias the review and the fixes. This skill runs the review and the feedback-addressing as two separate clean-context phases so neither is influenced by the current conversation. ## Core Features & Use Cases - Two-phase clean-context orchestration: Runs /code-review in a fresh-context subagent that posts its verdict to the PR, then runs /address-feedback to gather all feedback, apply fixes, push, and watch checks until green. - Copilot review gating: Probes whether GitHub Copilot review is available for the repo and runs a background poll so Copilot's comments are included before fixes begin, skipping silently when unavailable. - Untrusted input boundary: Treats PR comments, review bodies, and CI logs as data rather than instructions, surfacing suspected injection attempts verbatim and disabling auto-merge when untrusted input is present. - Use Case: After opening a PR with /create-pr, run this skill with --auto-merge to have it reviewed, fixed, and merged without further manual steps, or run it without the flag to approve each gate yourself. ## Quick Start Review and address all feedback on the pull request for my current branch, then merge it once checks are green.
