pr-classify
Skill10 starsChanged 19 days agoMITView on GitHub
Triage a PR as safe to merge, nice to review, or reject to merge in a fresh, unbiased context, then act on the verdict. Use after writing a PR body with /pr, or when asked whether a PR is safe or ready to merge.
In this repository25 other files it gives its agents.
- AGENTS.mdAGENTS.md
- agents-md-improverSkill
- ai-tutorialSkill
- deassembleSkill
- excalidraw-diagramSkill
- find-overlapping-skillsSkill
- first-principlesSkill
- humanizerSkill
- improve-skillsSkill
- local-artifactSkill
- loom-transcriptSkill
- manage-contextSkill
- organize-skill-mcpSkill
- pair-consultSkill
- pair-optimizeSkill
- pair-ratchetSkill
- pr-reportSkill
- scan-sessions-and-learnSkill
- simplify-pythonSkill
- skill-flowSkill
- slack-commSkill
- socratesSkill
- step-by-stepSkill
- sub-consultSkill
- to-dbmlSkill
The file
We split this skill into the rules it gives an agent
1 rule. About 0 are common, 1 is distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈1
Distinctive · estimate
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- could be automatic: a setting could enforce it
- Distinctive: line 33
Ask once, naming exactly what will happen: "Merge #N into dev, then promote dev → main?" with options both / dev only / don't merge.
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
pr-classify
The step after /pr: decide whether this PR can ship without a human reading it.
| Class | Meaning | Action |
| safe to merge | Confident it introduces no production issue | Ask to merge; on yes, merge (and promote dev → main if the repo uses that flow) |
| nice to review | Code-quality issues or possible minor bugs, nothing production-critical | Report findings; a human reviews before merge |
| reject to merge | A production-critical issue | Report what must be fixed; leave the PR unmerged |
Why a fresh context
The agent that wrote the code (and the /pr body) reads the diff through its intent and sees what it meant to write, not what it wrote. So the verdict always comes from a fresh subagent that has never seen this conversation. In Claude Code that means a new general-purpose agent, never a fork: a fork inherits this conversation and its bias.
Hand it only pointers: the PR (or branch + base), the repo path, and references/classifier.md as its brief. Keep your summary, your reasoning, and "what this change does" out of the prompt; it reads the code itself.
Steps
- Pin the target. Argument, else the current branch's PR. No PR yet → the subagent classifies the branch against its base, and the merge offer includes opening the PR with the
/prbody. - Dispatch the fresh subagent and wait for its report.
- Relay the verdict as returned. You may escalate it if you know something it couldn't (e.g. a known prod incident in that area); never downgrade it, since your read is the biased one.
- Act on the class.
Safe → merge
- Ask once, naming exactly what will happen: "Merge #N into dev, then promote dev → main?" with options both / dev only / don't merge.
- dev → main flow: detect it from repo docs, CI that gates
mainon adevhead, or the PR targetingdev. The promotion carries every commit on dev, not just this PR, so checkmain..devfirst: anything else riding along is unclassified; send it through the same fresh classifier or list it in the question. Reuse an open dev → main PR if one exists. Match the repo's past promotion merge method (usually a merge commit; squashing dev → main diverges the branches and conflicts the next promotion). Keep dev alive: no--delete-branchon the promotion. - Merge through the normal path with the repo's merge method. If branch protection blocks it, report and stop; an admin override defeats the review the protection exists for.
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.

