| name | qa-bug-report |
| description | Drafts a structured, reproducible bug report from behavior observed in the running app. Use after a tester reproduces a defect and wants a clean report another tester could file without a follow-up conversation. Day 2 bug-reporting block of the QA Foundational training. |
QA Bug Report
Core principle: AI drafts from evidence. Observed behavior beats a guessed repro. Ground every field in what actually happened, not what probably happens.
Input: behavior observed in the running app (using the playwright-cli skill, or a recorded walkthrough / screenshots).
Output: a structured bug report using template.md that another tester could file and reproduce cold.
Workflow
Pre-flight checks
Run these before starting any work:
playwright-cli --version
playwright-cli install --skills
node -e "console.log('bug-report-' + Date.now())"
If playwright-cli is not found, install it by running these two commands separately (chaining with && can fail on Windows):
npm install -g @playwright/cli@latest
playwright-cli install --skills
Use the output from step 3 as your -s= value for every playwright-cli call this run.
Evidence capture: you may save screenshots and walkthrough captures into the .playwright-cli/ directory (e.g. playwright-cli screenshot --filename=.playwright-cli/<name>.png). Embed these in the bug report and any HTML report generated from it.
To give visibility into the process go ahead and run a background command playwright-cli show so we can see what the agent is doing.
Bug Report Progress:
- [ ] Step 1: Observe and capture the real behavior
- [ ] Step 2: Establish expected vs actual
- [ ] Step 3: Draft the report from the template
- [ ] Step 4: Judge it - could someone file this without asking you a question?
Step 1: Observe and capture
Reproduce the bug against the running app and capture real evidence:
- The exact steps taken (URLs, fields, values entered).
- The actual result - real error text, real state, real selectors/values, a screenshot or trace if available.
Do not paraphrase from memory. If you did not observe a step, do not write it as if you did.
Step 2: Expected vs actual
- Expected - cite the source of the expectation (a ticket, an AC, a stated rule, or app behavior elsewhere). The oracle.
- Actual - what the app did, in concrete terms.
If the bug is that the app allows something it should block, say what rule is being violated and where that rule comes from.
Step 3: Draft from the template
Fill template.md: title, environment, preconditions, steps to reproduce, expected, actual, severity, priority, evidence. Keep steps numbered and literal - real data values, not "enter a date."
Step 4: Judge it
-> STOP. Read the report back and ask:
- Could another tester reproduce this with only what is written here?
- What did I assume that is not in the evidence? Mark or remove it.
- Is the severity/priority justified by impact, or guessed?
Present the report plus this judgment. The evidence-vs-assumption check is the skill.
Critical rules
- Every step and value comes from observed behavior, not assumption.
- Expected result cites its source (ticket, rule, or app behavior).
- Steps are literal and numbered with real data.
- Severity and priority are justified, not decorative.
Anti-patterns
- A repro with vague steps ("set up the data, then trigger the bug").
- Expected/actual that restate each other without a real oracle.
- Inventing error text or state that was not observed.
- Filing severity "critical" with no impact statement.
