agentleFS
marcin-urbanskimarcin-urbanski/claude-skills/skills/pr/SKILL.md

pr

Skill0 starsChanged 7 days agoMITView on GitHub

Format for a pull request body that a human can review fast - a small visual of the change, before-and-after evidence, and a merge danger call (one-way or two-way door, blast radius). Use whenever writing or rewriting a PR description, including at the end of spec-first.

The file

We split this skill into the rules it gives an agent

3 rules. About 0 are common, 3 are distinctive, and 0 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈3

    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
  1. Distinctive:

    Write the last three sections only when they have content:

    line 35
  2. Distinctive:

    Acceptance criteria: when there is a spec, each criterion with how it is met (test name or manual check).

    line 36
  3. Distinctive:

    If something could not be verified (no runner, no device), say so here.

    line 85

≈ “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

PR body

Adapted from Matt Pocock's pr skill (mattpocock/skills, MIT), whose Summary visuals come from Dex Horthy's show-me skill (humanlayer/skills, MIT).

The body tells the reviewer in under a minute what changed, whether it works, and how carefully to read it. Write it for a reviewed diff, from the diff, the spec and the test output. Start at ## Summary, with no preamble, and keep the prose brief.

## Summary

<one or two sentences: what and why>

<the smallest visual that makes the change clear>

## Evidence

- **Before:** <failing test, old output, or screenshot>
- **After:** <passing test, new output, or screenshot>

## Merge danger

**Door:** one-way | two-way. <why, one line>

**Blast radius:** <who or what breaks if this is wrong>

## Acceptance criteria
## Notes for the reviewer
## Follow-ups

Write the last three sections only when they have content:

  • Acceptance criteria: when there is a spec, each criterion with how it is met (test name or manual check).
  • Notes for the reviewer: reuse decisions, security notes, and review findings that were not fixed, with the reason.
  • Follow-ups: links to issues created during the work.

A section another skill requires (such as design-check's ## Design) goes after Merge danger.

Summary visual

Pick the one view that makes the key point clear; occasionally two. Keep only the calls, files, states and boundaries the reviewer needs.

  • Logic or an algorithm: pseudocode.
    on(save)
      if content is unchanged
        return cached result
      write new content
  • Runtime control flow: a call tree.
    submitOrder
      validateCart
      chargePayment
      sendConfirmation (queued)
  • UI structure: a component tree, with the state and module boundaries that matter.
  • A refactor or new files: a shallow file tree with one comment per entry.
  • Interaction between components or services: a Mermaid diagram (GitHub renders it).
    ```mermaid
    sequenceDiagram
        Admin->>App: apply 20% discount
        App->>Queue: DiscountApplied mail
        Queue-->>Customer: email
    ```
  • What changes inside a shape that already exists: a diff block of that tree or flow, not of the code.
     submitOrder
       validateCart
    +  applyDiscount
       chargePayment
  • Show a whole code block only when most of it is new and its shape is the point.

Evidence

Show a before and after, not "tests pass".

  • A screenshot is strongest for a visual change. gh cannot upload images, so use one only when the user provides it or the repo already hosts it.
  • Otherwise name the test that failed before and passes now, or show the command output that changed, trimmed to the lines that matter.
  • If something could not be verified (no runner, no device), say so here.

Merge danger

The door call tells the reviewer where to spend attention: skim a two-way door with a small blast radius, read a one-way door slowly.

  • Two-way door: a revert undoes it completely. Code-only changes, feature-flagged work, additive migrations.
  • One-way door: a revert does not undo the effect. Examples: a migration that drops or rewrites columns or data; emails, push notifications or webhooks sent to real people; queued or scheduled jobs that act on production data; payments, refunds, orders or stock (WooCommerce included); a public API or URL change that clients depend on; a Core Data or SwiftData model migration shipped to devices; an App Store submission.
  • Blast radius: everything that breaks if the change is wrong: which users, screens, integrations or data, and how many. Consider layout shifts, mobile, other consumers of a changed function or API.

You wrote the change, so you are biased toward calling it safe. When in doubt, call it one-way and say what makes it so.

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.