The Agent Skills Directory

[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection because it ingests untrusted data from external websites during SEO audits without explicit sanitization or boundary markers.
Ingestion points: Content retrieved from external URLs via tools like web_fetch or curl for technical and on-page analysis.
Boundary markers: The instructions lack specific delimiters or warnings to the agent to ignore embedded instructions within the fetched HTML or metadata.
Capability inventory: The agent is empowered to perform network operations (reads) and generate detailed reports based on those reads.
Sanitization: There are no instructions for filtering, escaping, or validating the content retrieved from audited sites before processing.

seo-audit