The Agent Skills Directory

[SAFE]: The skill instructions and reference materials (ai-writing-detection.md, international-seo.md) focus exclusively on SEO best practices and industry-standard documentation. No malicious command execution or suspicious network patterns were identified.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content from external websites to perform SEO audits.
Ingestion points: The agent retrieves HTML and metadata from user-specified URLs using web_fetch and curl as outlined in SKILL.md.
Boundary markers: The instructions provide a rigid 'Audit Report Structure' and 'Audit Framework' that effectively scope the agent's interactions with untrusted external data.
Capability inventory: The agent is authorized to fetch web content and read local project context files (.agents/product-marketing.md).
Sanitization: Audit steps specifically target SEO elements (robots.txt, canonical tags, heading structure), which limits the processing of unrelated or potentially malicious instructions embedded in the source HTML.

seo-audit