// Pull PR feedback from GitHub, Qodo, Copilot, and SonarCloud signals, then triage it into fixes, pushback, replies, resolutions, and follow-up issues. Use when a PR has review comments or failing checks and you want one repeatable workflow to gather and address them.
| name | pr-review |
| description | Pull PR feedback from GitHub, Qodo, Copilot, and SonarCloud signals, then triage it into fixes, pushback, replies, resolutions, and follow-up issues. Use when a PR has review comments or failing checks and you want one repeatable workflow to gather and address them. |
Use the helper scripts in scripts/ to pull feedback and manage inline review
threads. Treat gh as the source of truth for this workflow.
Before doing anything else:
gh --version works.gh auth status succeeds.$ARGUMENTS or from the current branch.Run:
python3 skills/pr-review/scripts/fetch_feedback.py --pr <number> --format markdown
This script gathers:
Treat the sources as:
Sort each finding into one of these buckets:
Deduplicate overlapping comments across Qodo, Copilot, and repeated inline threads before acting.
Implement the in-scope fixes first. Then re-run the most relevant checks.
Typical examples:
For deferred items, create a GitHub issue with:
Use this especially for:
Use:
python3 skills/pr-review/scripts/thread_actions.py reply-and-resolve \
--pr <number> --comment-id <comment_id> --body "Handled in <summary>"
for inline review comments that have been addressed.
For multiple inline comments at once, create a JSON file like:
[
{
"comment_id": 1234567890,
"body": "Handled by switching the published link to the GitHub README."
},
{
"comment_id": 1234567891,
"body": "Handled by replacing the broken site-relative SKILL.md link."
}
]
Then run:
python3 skills/pr-review/scripts/thread_actions.py reply-and-resolve-batch \
--pr <number> --file /tmp/pr-review-batch.json
Use:
gh pr comment <number> --body "Grouped response..."
for top-level bot comments that cannot be resolved as review threads.
When pushing back, be concrete. Reference the repo convention or platform requirement that makes the suggestion incorrect.
Answer questions about using Codex in a repository by reading the local guide references. Use when the user asks about AGENTS.md, sandbox and approvals, day-to-day Codex workflow, best practices, or the onboarding path in this repo. Also use when a user wants a grounded explanation based on the guide docs rather than a generic answer.
Audit a repository's Codex readiness and propose the next high-value improvements. Use when the user wants to review AGENTS.md, commands, verification, docs quality, CI, MCP usage, worktrees, or exercise ideas before deciding what to improve next.
Interactive onboarding guide for developers who are new to Codex or setting Codex up in a new repository. Use when the user wants a step-by-step walkthrough for getting started, writing or improving AGENTS.md, understanding sandbox and approvals, choosing a safe first task, or learning the inspect-plan-implement-verify workflow.