PR Description Prompt

When to use this

Opening a pull request. A good description makes review faster and gives you a record of why a change was made, which is worth more six months later than the diff itself.

The template

Write a pull request description from this diff.

Structure:

## What
One or two sentences on what changed, in terms of behaviour, not implementation.

## Why
The problem this solves. If there is an issue reference in the commits, link it.

## How
Only the parts of the approach a reviewer could not infer from the diff. Skip this if the diff is self explanatory.

## Risk
What could break. Which existing behaviour this touches. Anything needing attention on deploy, such as a migration, a config change, or an environment variable.

## Testing
What you verified and how. Say plainly what is not covered.

Rules:
- Describe effects, not file names. Reviewers can see which files changed.
- No filler. No "this PR aims to".
- If the change is breaking, say so at the top.
- If a migration or manual deploy step is needed, say so at the top.
- Do not claim anything was tested unless the diff shows a test for it.
- Keep it under 300 words unless the change genuinely warrants more.

Diff:
[paste here]

What to change

Match the section headings to your team's PR template if you have one.

If your repository uses conventional commits or a changelog generator, add a line asking for the correct prefix.

Remove the Testing section if your CI output already makes it obvious.

Why it is shaped this way

Putting Why before How is deliberate. The reason for a change is the part that gets lost, and the part a reviewer most needs in order to judge whether the approach fits.

The Risk section is the one people skip and reviewers most want. Even "no risk, this only touches the test suite" is useful, because it tells the reviewer how carefully to read.

The rule against claiming untested work matters here specifically. A model writing a PR description will otherwise produce a confident Testing section describing verification that never happened, and that misleads the reviewer in exactly the place they are trusting you.