What it does
A toolkit for testing local web apps through Playwright (Python scripts): interact with the UI, take screenshots, inspect the rendered DOM, read browser console logs. It ships a with_server.py helper that manages the server lifecycle (one process or several backend/frontend ones at once). The pattern: reach the app → wait for the JS to load → inspect the state → fire targeted clicks at the selectors you found. Dual nature: it overlaps with Playwright MCP, but here it is specifically an agent skill for "run this web flow end to end".
How to set it up
It ships in the example-skills bundle:
/plugin install example-skills@anthropic-agent-skills
Or drop the webapp-testing/ folder into ~/.claude/skills/. In Claude Code it is available as the webapp-testing skill.
When to use it
- Before closing out a web task — a mandatory browser pass: open the flow, click through it, check nothing next to it broke.
- Frontend debugging: catch a console error, screenshot the state, verify selectors.
Example
"Open the local build, go through login and checkout, screenshot every step" — the skill brings up the server, clicks through the flow with Playwright and hands back screenshots plus console errors.