# Github Pr Review Bot

> Use when operating a GitHub pull-request loop with the Codex review bot: triggering review, polling reactions/reviews/threads, addressing findings, or proving a clean current-head result.

- Skill: `applicate2628/github-pr-review-bot-2` (Agent Skill, multi-file: 2 files)
- Install (CLI): `npx skillmds@latest add applicate2628/github-pr-review-bot-2`
- Raw SKILL.md: https://api.skillmd.com/api/skills/applicate2628/github-pr-review-bot-2/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Coding & Dev Tools
- Author: applicate2628 (https://skillmd.com/u/applicate2628)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/applicate2628/github-pr-review-bot-2

---


# GitHub PR Review Bot

Drive one GitHub-hosted Codex review loop to a terminal result on the current remote head. This skill observes and coordinates the bot; it does not own code decisions, publication approval, CI, or merge.

## Binding contract

- Use hosted GitHub state, never local ancestry or remembered status.
- Bind every request and result to the current `headRefOid` and the exact `@codex review` comment ID.
- Trigger or resolve threads only with explicit user authorization or a standing authorization for that PR.
- Do not start or rerun CI as part of this loop.
- Never retrigger solely because time elapsed. A bot-authored `eyes` reaction on the exact trigger is acknowledged/in progress, not clean; keep polling that run.
- Verify reaction authors, not aggregate reaction counts.

## State machine

| Hosted evidence | State | Action |
| --- | --- | --- |
| Head changed after the trigger | stale | Ignore old terminal claims; trigger once on the new head when authorized. |
| Bot `eyes` on the exact trigger | in progress | Poll reviews, reactions, and current unresolved threads; do not retrigger. |
| Bot `+1` on the exact trigger, head unchanged, no unresolved current bot thread | clean | Record bot PASS; continue any separate human/publication gates. |
| Same-head bot review with no current bot finding comment and no unresolved current bot thread | clean | Record bot PASS; continue any separate human/publication gates. |
| Same-head bot review with current finding comments or an unresolved current bot thread | findings | Verify each finding, fix and test, push, then resolve only the exact fixed threads and trigger one new review. |
| No bot terminal output and no acknowledged run | indeterminate | Re-read authoritative state; do not invent a timeout or autonomous retrigger. |

## Hosted probes

Use `gh pr view <pr> --repo <owner>/<repo> --json headRefOid,state,isDraft,mergeable,updatedAt,url` first. Then query:

- exact issue comments and reaction actors through `gh api repos/<owner>/<repo>/issues/<pr>/comments` and `.../issues/comments/<id>/reactions`;
- submitted reviews through `gh api repos/<owner>/<repo>/pulls/<pr>/reviews`;
- `reviewThreads { id isResolved isOutdated path line comments }` through paginated GraphQL.

Anchor each poll on the latest hosted head and unfiltered latest bot state. Time windows are secondary only.

## Finding lifecycle

Treat bot text as a hypothesis until reproduced or verified in source/runtime. Do not bulk-resolve. After the exact fix is present on the hosted head, re-read its thread, resolve that thread, and leave unrelated or still-valid threads open. One new head gets at most one active review trigger.

Clean bot output is not merge or publication authority. Human review, leak scan, branch protection, and repository gates remain separate.

## Terms and Abbreviations

- **CI** — Continuous Integration.
- **PR** — Pull Request.
- **head SHA** — the hosted commit identifier in `headRefOid`.