Site skills built by watching a site in your own signed-in Chrome — network-first discovery, a surface survey or your own URL list, a skill roster for approval, then skills that survive UI changes, keep the session alive, clean up after themselves and probe for drift. Semantic handles only. Every observation is classified publishable, private, or never-recorded.
1 on GitHubby Daniel Rosehillv1.2.0MITupdated 18 days ago
Not in a marketplace yet, so it loads for one session with --plugin-dir.
Write an approved site skill so it survives small UI changes and fails loudly when it does not — handles, assertions, provenance.
close-out-session
End a browser-driven build or skill run cleanly — stop keep-alives, close only the tabs the agent opened, confirm what was committed, report to the user.
drift-probe
Re-verify a site plugin's load-bearing routes and handles, and report CHANGED with what moved — before a user hits it as a wrong answer.
map-site-api
Map the JSON endpoints a site's own pages call, so skills can read data from a stable endpoint instead of scraping rendered markup.
observe-page
Observe one page or a list of pages in the browser and write an observation note — what data is there and which durable handles reach it.
pick-browser-route
Decide how to reach a site — a signed-in browser the user can watch (a persistent agent browser if they have one, else their own Chrome via claude-in-chrome), headless Playwright for public pages, plain fetch when it works.
propose-site-skills
Turn observations of a site into a ranked roster of candidate skills for the user to approve or reject before any are written.
publication-gate
Classify every observation as publishable, private, or never-recorded before committing — plus the stop rule for security findings.
repair-broken-skill
Fix a site skill that broke when the site changed — re-observe the surface, move the handle up the ladder, and record what moved.
session-keepalive
Detect a signed-in session timing out mid-build or mid-run, and keep it alive with a bounded periodic ping only while the agent is actively driving it.
site-recon-intake
Start building a plugin for a website — capture the target, the page list, the account posture, and what the skills must do. Run this first.
survey-site-surface
Exploratory brief — walk a signed-in site's navigation and network calls and propose which pages are worth wrapping in skills, before observing any in depth.
write-site-profile
Write or update the site profile JSON — the one file holding every volatile fact about a site, so a redesign is a one-file edit.
Slash commands (2)
/site-skill-builder:site-drift
Re-verify a site plugin against the live site and repair what the site has changed underneath it.
/site-skill-builder:site-plugin
Build or extend a Claude Code plugin for a website by observing it in the browser — intake, observe, propose, approve, author.
Subagents (2)
observation-recorder
Turn captured page material into a structured observation note — handles, assertions and traps, inventing nothing.
skill-durability-auditor
Audit a drafted site SKILL.md against the durability doctrine and report brittle handles, missing assertions, and silent-failure paths.
Web scraping, Google search, structured data extraction, and MCP server integration powered by Bright Data. Includes 7 skills: scrape any webpage as markdown (with bot detection/CAPTCHA bypass), search Google with structured JSON results, extract data from 40+ websites (Amazon, LinkedIn, Instagram, TikTok, YouTube, and more), orchestrate Bright Data's 60+ MCP tools, built-in best practices for Web Unlocker, SERP API, Web Scraper API, and Browser API, Python SDK best practices for the brightda...
Web scraping, Google search, structured data extraction, and MCP server integration powered by Bright Data. Includes 7 skills: scrape any webpage as markdown (with bot detection/CAPTCHA bypass), search Google with structured JSON results, extract data from 40+ websites (Amazon, LinkedIn, Instagram, TikTok, YouTube, and more), orchestrate Bright Data's 60+ MCP tools, built-in best practices for Web Unlocker, SERP API, Web Scraper API, and Browser API, Python SDK best practices for the brightda...
Browser automation in Claude Code through Skyvern Cloud: open pages, fill forms, extract structured data, take screenshots, and run repeatable workflows. Sign in with OAuth from /mcp. Runs can use Skyvern credits.
Agent Skill for general-purpose browser automation with Playwright. The agent writes and executes custom scripts for testing pages, validating UX, and automating browser tasks.
1 skill
lackeyjb/playwright-skill 3,288
Maintain site-skill-builder?
This page updates itself from your repo. Add the badge to your README, or get it Featured to pin it above the organic results with a followed link to your site.