Diff-aware browser QA

Derive a 5-15 check test plan from a diff, run it in a real browser, and report issues with evidence.

Run it as a prompt

Paste this into any AI agent, or fetch it: curl -s https://uplift.page/api/v1/prompts/diff-aware-browser-qa/raw

prompt.md

# Diff-aware browser QA

Test what this change actually affects, in a real browser, and report problems with evidence.

## Context
- Changed files, diff, or branch: [DIFF OR BRANCH]
- App URL (local or staging): [APP URL]
- Login: [TEST ACCOUNT OR NONE]. Use only a test account from the project's seed or example config, never a personal or production account.

## Steps
1. From the diff, list the user-facing pages and API routes affected, plus neighbouring pages that share components or data with them.
2. Write a focused plan of 5-15 checks: the main flow, one empty state, one error state, one permission check, and one check at 375px width.
3. Run the plan with a browser automation tool (for example Playwright). For each check: navigate, interact (click, fill, submit), save a screenshot as qa/<check-name>.png, and record console errors and failed network requests (4xx/5xx).
4. Re-run each failure once to rule out flakiness, and note if it is intermittent.
5. Spot-check 2-3 neighbouring areas for regressions.

## Rules
- Report only what you observed. Label any guess about the cause as a guess.
- Do not submit forms that send real email, charge cards, or touch production data. Skip them and say so.
- Do not change application code during QA.

## Output
1. Coverage: pages and routes tested, and anything skipped with the reason.
2. Issues ranked critical/high/medium/low, each with repro steps, expected vs actual, screenshot file name, console or network evidence, and a suspected cause.
3. Ship recommendation: ready / ready with known issues / not ready, in one sentence.

The full prompt

Test what this change actually affects, in a real browser, and report problems with evidence.

Context

Changed files, diff, or branch: [DIFF OR BRANCH]
App URL (local or staging): [APP URL]
Login: [TEST ACCOUNT OR NONE]. Use only a test account from the project's seed or example config, never a personal or production account.

Steps

From the diff, list the user-facing pages and API routes affected, plus neighbouring pages that share components or data with them.
Write a focused plan of 5-15 checks: the main flow, one empty state, one error state, one permission check, and one check at 375px width.
Run the plan with a browser automation tool (for example Playwright). For each check: navigate, interact (click, fill, submit), save a screenshot as qa/.png, and record console errors and failed network requests (4xx/5xx).
Re-run each failure once to rule out flakiness, and note if it is intermittent.
Spot-check 2-3 neighbouring areas for regressions.

Rules

Report only what you observed. Label any guess about the cause as a guess.
Do not submit forms that send real email, charge cards, or touch production data. Skip them and say so.
Do not change application code during QA.

Output

Coverage: pages and routes tested, and anything skipped with the reason.
Issues ranked critical/high/medium/low, each with repro steps, expected vs actual, screenshot file name, console or network evidence, and a suspected cause.
Ship recommendation: ready / ready with known issues / not ready, in one sentence.

Served from the uplift.page library and refreshed within 5 minutes of every update.