skill-pr
Skill0 starsChanged 10 days agoMITView on GitHub
Draft and preflight one water-skills contribution against the current repository guidance and pull request template. Use when preparing or reviewing a contribution or PR description.
In this repository9 other files it gives its agents.
- AGENTS.mdAGENTS.md
- epanet-validationSkill
- eval-writingSkill
- pywr-engineeringSkill
- scientific-writingSkill
- skill-browserSkill
- skill-evaluationSkill
- skill-writingSkill
- wntr-engineeringSkill
The file
We split this skill into the rules it gives an agent
2 rules. About 0 are common, 2 are distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈2
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
- Distinctive: line 20
Never invent an incident, transcript, prior-PR search result, test or evaluation outcome, model or harness version, installed plugin, …
- Distinctive: line 21
Do not mark a human-review checkbox or claim approval until the human confirms review of the complete proposed diff.
≈ “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
Skill PR
Prepare a truthful draft and preflight for one contribution. Keep the work in the current checkout and return the draft for the human to review and submit.
Procedure
- Read current guidance. For each request, read
README.md,CONTRIBUTING.md, and.github/PULL_REQUEST_TEMPLATE.mdfrom the current checkout. Use the template as the draft structure,CONTRIBUTING.mdfor contributor requirements, and the README as project context. For PR procedure, follow explicit requirements in the template andCONTRIBUTING.md; report any conflict with README rather than silently reconciling it. Do not treatAGENTS_obsolete.mdas authority or assume a missing referenced document is available. - Bound one contribution. Inspect the proposed change and its read-only working-tree diff when available. Establish the single problem and the files intended for this contribution. If the intended scope is unclear or includes unrelated work, ask the human to identify the contribution before drafting; do not stage, edit, or discard changes as part of preflight.
- Collect evidence. Compare the contribution with every current template heading, field, checkbox, instruction, and applicable conditional section. Ask concise, specific questions for missing facts. In particular, obtain evidence for the experienced problem, project fit, alternatives, related open and closed PRs, authoring and test environments, validation and evaluation results, and human review wherever the current template requires them. Use only facts supplied by the human or directly verified in the available checkout or runtime. If a read-only PR search is unavailable, ask the human to search open and closed PRs and provide the result.
- Preflight the contribution. Check that it addresses one concrete problem, fits the repository's stated scope, and accounts for the current contribution guidance and template. Run a relevant existing validation only when appropriate and already available; report the exact command and observed result. Do not install dependencies. Mark checks that were not run or could not be verified as pending.
- Draft and report status. Follow the current template's section order and account for every section. Fill only evidence-backed answers. For unanswered fields, use a clearly labeled
[NEEDS HUMAN INPUT: …]marker in a partial draft and identify the specific information needed; never imply the draft is complete. Report preflight findings and blockers separately from the draft.
Evidence and handoffs
- Never invent an incident, transcript, prior-PR search result, test or evaluation outcome, model or harness version, installed plugin, author, reviewer, or approval. Distinguish what was observed from what remains unverified.
- Do not mark a human-review checkbox or claim approval until the human confirms review of the complete proposed diff. Hand the complete diff and any unresolved evidence requests to the human for review.
- The human handles any unavailable prior-PR search, supplies missing evidence, confirms the final text and target, and submits the PR themselves.
- This skill stops at drafting and preflight. It never creates branches, commits, pushes, or submits pull requests.
Completion criteria
Drafting and preflight are complete when exactly one contribution is in scope, every current template section is represented, factual answers are supported by evidence, missing evidence and unrun checks are explicit, and the human handoffs are clear. Report the draft as not ready while required information or review is outstanding; only the human decides whether to submit.
Evaluation status: This skill's behavior has not been evaluated.
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.

