PostHog Debugger

PostHog's own skill for checking how PostHog is set up on a live website. Give your agent a URL and it opens the page with PostHog's debug mode on, reads the posthog object, the console and the network traffic, and writes up a structured summary of how PostHog is installed and what it's sending. It only looks; it never changes anything. What you can do with it - See how PostHog is loaded: on the window, bundled into the app or through a custom proxy, from PostHog's CDN or self-hosted - Check the library version, the API host (US or EU cloud, a proxy or self-hosted), the project token and the persistence type - See which features are loaded and which are actually enabled: autocapture, session recording, surveys, heatmaps, feature flags and web vitals - List the feature flags loaded on the page with their values, plus the distinct and session IDs - Confirm that events, recordings and flag requests are reaching PostHog - Catch common problems: more than one PostHog instance, an opted-out user, debug mode left on in production, autocapture turned off - Spot the other analytics and tracking tools on the page How it works It's a single Agent Skill, packaged as a Claude Code plugin in PostHog's skills marketplace (PostHog/skills). It drives a browser through Chrome DevTools MCP (PostHog's recommendation) or Playwright MCP, so add one of those to your agent first; Chrome DevTools MCP needs Chrome running with remote debugging on. For a page behind a login, it opens the page and asks you to sign in yourself in that browser, then carries on. Good to know It comes from PostHog's customer success team and sits with the team skills in PostHog's skills repo, which PostHog marks as under construction. It doesn't call PostHog's API, so it needs no PostHog key: it reads only what the page itself loads.