The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes the chrome-devtools-mcp suite (navigate_page, evaluate_script) to automate browser interactions with pagespeed.web.dev. This is the intended functionality for obtaining Lighthouse reports.
[INDIRECT_PROMPT_INJECTION]: The skill ingests data from external web pages by reading global JavaScript variables (window.__LIGHTHOUSE_MOBILE_JSON__). This represents an external data ingestion surface. However, the risk is minimal as the data is sourced from a trusted Google domain and the extraction logic specifically targets structured performance metrics rather than executing untrusted content.
Ingestion points: window.__LIGHTHOUSE_MOBILE_JSON__ and window.__LIGHTHOUSE_DESKTOP_JSON__ read via evaluate_script in SKILL.md.
Boundary markers: Not explicitly defined in scripts, but the extraction logic uses fixed keys (e.g., lhr.audits['speed-index']).
Capability inventory: navigate_page, evaluate_script, new_page, list_pages (used for browser control).
Sanitization: The skill uses a JS mapping function to filter and extract only relevant audit IDs and scores, ignoring the rest of the object's content.

pagespeed-insights