agentleFS
Sequela02Sequela02/canasta/skills/workflow/writing-pr/SKILL.md

writing-pr

Skill0 starsChanged yesterdayMITView on GitHub

Write or refresh a pull request title and body from the current

1 other project uses a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

6 rules. About 0 are common, 4 are distinctive, and 2 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈4

    Distinctive · estimate

    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
  1. Very specific:

    Use for PR titles and bodies.

    line 14
  2. Distinctive:

    Before writing anything, read the repository's contribution rules, its PR template and two recent merged PRs, and establish whose …

    line 18
  3. Distinctive:

    Write the title and opening around the concrete problem and resulting behavior.

    line 29
  4. Distinctive:

    Use the smallest useful representation from technical-writing.

    line 40
  5. Very specific:

    Use benchmark-readiness-review for claim validity.

    line 52
  6. Distinctive:

    When the repository or your own organization requires a risk note, end the PR with it: one-way or two-way door, blast radius, and evidence.

    line 66

≈ “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.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Writing Pull Requests

Use for PR titles and bodies. Use technical-writing for prose and visual explanations. This procedure describes delivered work; writing-issues describes the problem to solve. Neither owns branch, publication, or merge authority.

  1. Before writing anything, read the repository's contribution rules, its PR template and two recent merged PRs, and establish whose repository it is from its remote and maintainers, not from assumption. Establish the intended base and current candidate. Inspect the aggregate diff, linked issue or accepted objective, repository PR template, and contribution rules. Check whether a PR for the branch already exists, and inspect its metadata when updating. Verify claims inherited from commits or previous summaries; report unavailable material evidence. The repository's template and the structure of its recent merged PRs decide the sections. In a project you don't maintain, they replace your own organization's conventions; use upstream-contribution there too.
  2. Write the title and opening around the concrete problem and resulting behavior. A reviewer unfamiliar with the conversation should understand why this change exists. Follow repository naming conventions; recent merged PRs and history show them, and a PR title often becomes the merge commit. Name the effect that matters, not the mechanism. Do not promise unverified benefits in the title. Describe the final aggregate change, not intermediate attempts, session history, or transient line counts.
  3. Explain mechanisms, meaningful tradeoffs, compatibility changes, and review focus only where the diff does not make their implications clear. Name affected boundaries and give a reading order when it materially helps. Do not invent rejected alternatives or narrate each changed file.
  4. Use the smallest useful representation from technical-writing. For UI changes, prefer comparable real before/after captures; use a short recording when interaction or timing matters. Identify the states and revisions shown. Missing captures stay missing, never replaced by generated images presented as proof. Narration and diagrams explain behavior; they do not verify it.
  5. Include relevant observed verification and material gaps once. Distinguish passed, failed, unrun, and pending checks; collect executable evidence with tdd and its end-to-end verification reference when it is missing. Never claim TDD from passing tests alone: a red/green claim needs an observed relevant failure and subsequent pass. Skip test inventories and generic claims that everything works.
  6. For performance claims, show the actual baseline and candidate revisions, workload, conditions, units, and measurement uncertainty with the comparison. Use benchmark-readiness-review for claim validity. Missing comparable results mean the improvement is unverified; never fabricate a before/after table.
  7. Preserve unresolved findings, meaningful residual risks, and required evidence even when concise. Respect repository templates; otherwise scale from a short paragraph and verification line to a structured explanation for complex work. Omit empty sections, repeated facts, raw metadata dumps, and ceremonial lists.
  8. Reconcile the whole title and body with the current candidate after scope or implementation changes. Check every linked artifact exists and is suitable for its audience. Before publishing, check the title, body, quotes and attachments for secrets, private names and internal infrastructure (secret-store names and layout, private repositories, hosts). Return the draft or use the already authorized publication workflow (a real PR, not a draft, unless the repository says otherwise); do not add approval gates or infer authority from the skill.
  9. When the repository or your own organization requires a risk note, end the PR with it: one-way or two-way door, blast radius, and evidence. A PR to another project carries none unless that project asks for one. The agent never merges unless the project's stated policy allows it; passing checks and drafting a PR do not grant merge authority.

Download now (ZIP)

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
What did you use it for? e.g. Reviewing pull requests in a Rails app

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.