菜单
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.
默认使用会先检查来源的 Prompt；你也可以切换为直接命令，或下载本地副本。
决定是否安装前，请先阅读 SKILL.md，以及 SkillsMP 当前展示的配套文件。
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
直接命令不会经过审查 Prompt；运行前请先检查来源。
npx skills add https://github.com/devclarityai/qa-environment-setup --skill qa-bug-report命令会保持在同一行。复制前请横向滚动并检查完整内容。
想先保存到本地？可下载 SkillsMP 当前能够提供的文件。
Verify and repair the local environment for Playwright-based QA training - node and npm, repo dependencies, browser binaries, playwright-cli, the Playwright MCP server, and .playwright.env - then prove it works with one smoke test. Covers macOS, Windows (PowerShell), and WSL.
Guides the creation of Playwright API tests from exploration through implementation. Covers API discovery, test plan creation, helper/data factory development, and test writing with human checkpoints at each stage.
Analyze a test automation codebase and produce a single-page HTML "model of what exists" report covering assertion depth, test levels, run mechanics, environments, test data, authentication, network handling, dependencies and security, run commands, locators and consistency, and AI tooling artifacts.
正在显示 SKILL.md
基于 SOC 职业分类
| 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. |
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.
Run these before starting any work:
# 1. Verify playwright-cli is installed
playwright-cli --version
# 2. Ensure playwright-cli skills are installed
playwright-cli install --skills
# 3. Session name — use -s=<output> for all playwright-cli calls this run
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?
Reproduce the bug against the running app and capture real evidence:
Do not paraphrase from memory. If you did not observe a step, do not write it as if you did.
If the bug is that the app allows something it should block, say what rule is being violated and where that rule comes from.
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."
-> STOP. Read the report back and ask:
Present the report plus this judgment. The evidence-vs-assumption check is the skill.