watch-pr
Skill118 starsChanged 37 days agoMITView on GitHub
Wait for a pull request's checks to settle, then report what passed and what failed. Blocks in one call rather than polling. A wait then costs one turn instead of one per look. Use this whenever the user asks to watch a pull request or wait on CI, and whenever a workflow needs that answer before it can go on.
10 other projects use a version of this.
Adopted by
Most recently taken up by
- spencerldixon/skillstheir own version, first seen Nov 8, 2026
- fitlab-ai/agent-infratheir own version, first seen Oct 27, 2026
- s3s-project/s3stheir own version, first seen Oct 8, 2026
- parawanderer/OpenTagViewertheir own version, first seen Oct 8, 2026
- ayutaz/piper-plustheir own version, first seen Oct 8, 2026
- meta-pytorch/OpenEnvtheir own version, first seen Oct 8, 2026
Other versions10 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 10
- meta-pytorch/OpenEnv★ 2.9ksame name
- huggingface/OpenEnv★ 2.9ksame name
- parawanderer/OpenTagViewer★ 455same name
- s3s-project/s3s★ 315same name
- ayutaz/piper-plus★ 240same name
- fitlab-ai/agent-infra★ 88same name
- samsantosb/ship-it★ 37same name
- sherifabdlnaby/skills★ 9same name
- spencerldixon/skills★ 4same name
- rrjoson/claude-skills★ 0same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository32 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- analyze-sessionSkill
- codex-analyze-sessionSkill
- codex-commitSkill
- codex-fix-proseSkill
- codex-fix-prSkill
- codex-merge-prSkill
- codex-prSkill
- codex-rebaseSkill
- codex-resolve-rebase-conflictsSkill
- codex-review-commit-messageSkill
- codex-review-pr-descriptionSkill
- codex-review-prose-fixSkill
- codex-review-squash-messageSkill
- codex-watch-prSkill
- codex-write-pr-descriptionSkill
- codex-write-prose-fixSkill
- commitSkill
- fix-proseSkill
- fix-prSkill
- merge-prSkill
- prSkill
- rebaseSkill
- releaseSkill
- resolve-rebase-conflictsSkill
- review-commit-messageSkill
- review-pr-descriptionSkill
- review-prose-fixSkill
- review-squash-messageSkill
- write-pr-descriptionSkill
- write-prose-fixSkill
The file
We split this skill into the rules it gives an agent
2 rules. 0 are common, 0 are distinctive, and 2 are very specific to this project.
0
Common
Most of the 8 edited versions we read keep these.
0
Distinctive
This author's own choices. Worth a look.
2
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 16
Wait for the checks on a pull request to settle, then say what happened.
- Very specific: line 41
Never poll gh pr checks in a loop of your own.
“Common” is measured: kept by at least 6 in 10 of the 8 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Watch a pull request
Wait for the checks on a pull request to settle, then say what happened. This skill ends when the checks stop moving. Fixing what failed belongs to fix-pr, and merging belongs to merge-pr.
A guard hook runs alongside and refuses the watching forms this skill wraps, so a hand-rolled poll can't quietly replace the watcher. A plain gh pr checks for a one-off look remains allowed.
Which pull request
$ARGUMENTS names the pull request number when the caller knows it. An empty value means whatever is open for the current branch.
Step 1: watch
Arm the watcher through the Monitor tool, which turns each line the script prints into a notification:
Monitor({
command: "bash .claude/skills/watch-pr/scripts/watch-checks.sh <number>",
description: "checks on pull request <number>",
timeout_ms: 1800000,
persistent: false,
})Omit the number to use the current branch's open pull request.
Monitor is the right tool rather than a blocking call, because the script emits one line per check the moment that check settles. A failure arrives while the slower jobs are still going. The diagnosis then starts earlier and no turn goes to waiting. The script exits once nothing is pending, which ends the watch by itself.
Never poll gh pr checks in a loop of your own. Each look costs a turn and reports a state that has already moved on.
The lines to expect:
WATCHING #<number>once, at the startPASS <job>,SKIP <job>, orFAIL <job> <run link>, one per check as it settlesALL GREEN #<number>,FAILED #<number> with N failing check(s), orTIMEOUT #<number>to close
Running the same command through Bash also works and blocks until the same ending, with the exit code reporting the outcome: 0 all passed, 1 something failed, 2 the wait ran out. Prefer Monitor. Use the blocking form only where the next step can't start until the answer arrives.
Override the bounds through the environment when a run is unusually slow. PR_CHECKS_TIMEOUT sets the whole wait and PR_CHECKS_INTERVAL the poll spacing, both in seconds. Keep timeout_ms past PR_CHECKS_TIMEOUT so the script reports its own timeout rather than dying at the tool's.
Step 2: report
Say which checks failed and hand back. Give the pull request, the failing jobs, and the run links the script printed.
Where the caller asked only to watch, stop here. Where the caller asked for a fix, invoke fix-pr with the same number. That skill reads the logs and identifies the local task reproducing each failure.
A timeout isn't a failure. Say the checks were still running, give the elapsed bound, and let the caller decide whether to wait again.
Preconditions
ghinstalled and authenticated- a pull request open for the branch, or a number in
$ARGUMENTS
The script checks both and stops with a plain message when either is missing.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

