babysit-pr
Babysit a GitHub pull request after creation by continuously polling CI checks/workflow runs, new review comments, and mergeability state until the PR is ready to merge (or merged/closed). Diagnose failures, retry likely flaky failures up to 3 times, auto-fix/push branch-related issues when appropriate, and stop only when user help is required (for example CI infrastructure issues, exhausted flaky retries, or ambiguous/blocking situations). Use when the user asks Codex to monitor a PR, watch CI, handle review comments, or keep an eye on failures and feedback on an open PR.
92
Quality
92%
Does it follow best practices?
Impact
85%
2.12xAverage score across 3 eval scenarios
Evaluation results
56%
56%Automate the PR Babysitting Workflow
PR monitoring automation script
Continuous watch mode
0%
100%
Correct script path
0%
100%
Actions list dispatch
0%
100%
Terminal stop actions
0%
100%
Idle does not stop
0%
100%
Review before retry
0%
100%
CI fix commit message
0%
0%
Review fix commit message
0%
0%
Restart watch after push
0%
50%
Green-state exponential backoff
0%
0%
1-hour polling cap
0%
0%
1-minute non-green polling
0%
0%
100%
3%CI Failure Classification Guide
CI failure classification and retry commands
DNS/registry timeout = flaky
100%
100%
Runner startup failure = flaky
100%
100%
TypeScript error in touched file = branch-related
100%
100%
Snapshot mismatch in touched component = branch-related
100%
100%
External API rate limit = flaky
100%
100%
Log inspection command
100%
100%
Retry command uses --failed flag
100%
100%
3-retry limit documented
100%
100%
Branch-related action: fix and push
100%
100%
Ambiguous case: diagnose once first
70%
100%
100%
77%PR Monitoring Runbook
Polling cadence and stop conditions
1-min non-green polling
0%
100%
Exponential backoff when green
0%
100%
1-hour green-state cap
0%
100%
Reset interval on change
60%
100%
Three terminal stop conditions
30%
100%
Idle does not stop
0%
100%
Review before retry priority
0%
100%
Restart watch after push
40%
100%
Single watch process
100%
100%
Commit message formats
0%
100%
- Evaluated
- Agent
- Claude Code
- Model
- Claude Sonnet 4.6
Table of Contents
Is this your skill?
If you maintain this skill, you can claim it as your own. Once claimed, you can manage eval scenarios, bundle related skills, attach documentation or rules, and ensure cross-agent compatibility.