The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes content and metadata from external websites via Lighthouse and CrUX reports. This creates a surface where a maliciously crafted website could attempt to influence the agent's performance analysis.
Ingestion points: The url parameter in the seo_run_report tool, which fetches data from external web servers.
Boundary markers: The instructions provide context on how to interpret structured fields but do not define explicit boundary markers for the data content itself.
Capability inventory: The skill utilizes seo_list_reports, seo_describe_report, and seo_run_report to fetch and display site diagnostics.
Sanitization: There is no mention of sanitization or filtering of the content retrieved from the audited URLs before it is presented to the agent.

performance-audit