菜单
Browser automation with Playwright. Use when testing websites, taking screenshots, checking responsive design, filling forms, testing login flows, checking links, or any browser-based task. Auto-detects dev servers and writes scripts to .tmp/.
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。
基于 SOC 职业分类
Build data processing pipelines using DuckDB, GDAL, and gpio as pixi tasks. Use when creating ETL workflows, multi-step data transformations, format conversions, or batch processing — even if the user just says "process this data" or "build a pipeline".
Attach a DuckDB database for persistent querying. Use when the user has a .duckdb file they want to explore, or when they want to persist query results across sessions. Explores schema and wires up state.sql so all DuckDB skills can access it.
Search DuckDB and DuckLake documentation via full-text search. Use when encountering DuckDB errors, looking up function syntax, or when other skills hit unknown DuckDB behavior. Also use proactively when unsure about DuckDB features.
Install or update DuckDB extensions. Use when DuckDB reports a missing extension, when the user asks to install/update extensions, or when other skills delegate here. Each argument is a plain extension name or name@repo. Pass --update to update.
Run SQL queries against DuckDB databases or ad-hoc files. Accepts raw SQL or natural language. Use whenever the user wants to query data, explore tables, run analytics, or ask questions about datasets — even if they don't say "DuckDB" explicitly.
Read and explore any data file (CSV, JSON, Parquet, Avro, Excel, spatial, remote) using DuckDB. Use proactively when the user mentions a data file, asks "what's in this file", or wants to explore/preview/describe any dataset — even without saying DuckDB.
| name | playwright-skill |
| description | Browser automation with Playwright. Use when testing websites, taking screenshots, checking responsive design, filling forms, testing login flows, checking links, or any browser-based task. Auto-detects dev servers and writes scripts to .tmp/. |
| allowed-tools | Bash, Read, Write, Glob |
Skill dir: .claude/skills/playwright-skill
Temp dir: .tmp/ (gitignored, all scripts/screenshots go here)
Env vars (inside pixi run): PIXI_PROJECT_ROOT, INIT_CWD, PIXI_PROJECT_NAME
pixi run node --version
mkdir -p .tmp
cd .claude/skills/playwright-skill && pixi run pnpm run setup
Detect dev servers:
cd .claude/skills/playwright-skill && pixi run node -e "require('./lib/helpers').detectDevServers().then(s => console.log(JSON.stringify(s)))"
1 server → use it. Multiple → ask user. None → ask for URL.
Write script to .tmp/playwright-test-*.js (never to skill dir)
Execute:
pixi run node .claude/skills/playwright-skill/run.js .tmp/my-script.js
const { chromium } = require('playwright');
const path = require('path');
const TARGET_URL = 'http://localhost:3001';
const TMP = path.join(process.env.PIXI_PROJECT_ROOT || process.cwd(), '.tmp');
(async () => {
const browser = await chromium.launch({ headless: false });
const page = await browser.newPage();
await page.goto(TARGET_URL);
// Your automation here...
await page.screenshot({ path: path.join(TMP, 'result.png'), fullPage: true });
await browser.close();
})();
[{w:1920,h:1080}, {w:768,h:1024}, {w:375,h:667}], screenshot eachpage.fill('input[name="email"]', ...) → page.click('button[type="submit"]') → page.waitForURLwaitForSelector('.success')page.locator('a[href^="http"]').all() → HEAD request eachconst helpers = require('./lib/helpers');
await helpers.safeClick(page, 'button.submit', { retries: 3 });
await helpers.safeType(page, '#username', 'testuser');
await helpers.takeScreenshot(page, 'result');
await helpers.handleCookieBanner(page);
const data = await helpers.extractTableData(page, 'table.results');
PW_HEADER_NAME=X-Test PW_HEADER_VALUE=true pixi run node .claude/skills/playwright-skill/run.js .tmp/test.js
See API_REFERENCE.md for selectors, network interception, auth, visual regression, mobile emulation.