Pull Request Quality Contract

/pull-request-quality-contract

Use this repository-local skill before completing work or opening or updating a pull request. It keeps the review evidence concrete: why the change exists, what changed, what proof exists, which commands ran, and which checks were deliberately skipped.

Usage

/pull-request-quality-contract

What this skill does

1. Reads .agents/DEFINITION_OF_DONE.md and .github/pull_request_template.md.
2. Maps the changed behaviour to the smallest relevant proof: tests, command output, screenshots, video, or a technical Not applicable reason.
3. Checks that skipped proof names the reason, risk, and follow-up owner.
4. Writes or updates the pull request body using the repository template.

Implementation steps

1. Read .agents/DEFINITION_OF_DONE.md before implementation starts. Completion criterion: the pull request scope, required proof, and verification commands are known before coding.
2. Map behaviour changes to tests. Completion criterion: every user-visible behaviour changed by the pull request has an automated behaviour or end-to-end test, or an explicit technical reason it cannot be automated.
3. Produce behavioural proof for user interface changes. Completion criterion: the pull request contains current video and screenshots from the changed branch, or states Not applicable with the reason.
4. Run verification on the final branch. Completion criterion: format, lint, type, unit, integration, and end-to-end checks relevant to the changed files have been run after the last code change.
5. Write the pull request body using .github/pull_request_template.md exactly. Completion criterion: the body contains, in order, Why does this feature exist?, What changed?, Behavioural Proof (with video and screenshots), and Verification Summary.
6. Do not mark work complete with placeholders. Completion criterion: every proof link, screenshot path, video path, test command, failure, skipped check, and residual risk is explicit.

Evidence Rules

• Test claims require command output.
• User interface behaviour claims require screenshots or video from the branch under review.
• End-to-end claims require named scenarios and their pass/fail result.
• Skipped proof requires a technical reason, not convenience.
• Existing unrelated failures must be separated from failures introduced by the pull request.

What this skill explicitly does NOT do

• It does not invent proof that was not produced.
• It does not require screenshots or video for non-visual documentation, governance, or script-only changes.
• It does not replace the repository validator or continuous integration checks.
• It does not mark a pull request ready while relevant checks are still failing, queued, or running.

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: BenSheridanEdwards
• Source: BenSheridanEdwards/ArchitectPlaybook
• License: MIT

Install and usage instructions live in the source repository linked above.