lighthouse
Pass
Audited by Gen Agent Trust Hub on Sep 6, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADSDYNAMIC_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes content from external URLs and audit reports. Maliciously crafted web pages could attempt to influence agent behavior via report text. The skill mitigates this by instructing the agent to treat all page content and report descriptions as data rather than instructions.
- Ingestion points: Reads data from external URLs and supplied JSON/HTML report files (SKILL.md, references/reports-and-improvements.md).
- Boundary markers: Explicitly uses warnings to "Keep the evidence boundary" and treat data as non-executable (SKILL.md).
- Capability inventory: Includes file system writes, browser automation via Puppeteer, and CLI command execution.
- Sanitization: Instructs the user to inspect and sanitize reports before sharing to remove authentication headers or sensitive data.
- [COMMAND_EXECUTION]: The skill executes various CLI tools including
lighthouse,lhci, andnodeto perform audits and run helper scripts. - [DYNAMIC_EXECUTION]: Includes two Node.js scripts (
audit-flow.mjsandsummarize-lighthouse.mjs) used to automate browser interactions via Puppeteer and parse complex JSON audit results at runtime. - [EXTERNAL_DOWNLOADS]: References and fetches documentation and configuration from Google Chrome's official GitHub repositories.
Audit Metadata