The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to crawl and ingest untrusted data from external websites as part of its core SEO auditing functionality. This surface is well-managed; instructions guide the agent to extract specific technical attributes (e.g., meta tags, canonicals, JSON-LD) and report them as evidence, rather than executing or following natural language instructions encountered in the crawled content.
[COMMAND_EXECUTION]: The skill performs network operations to fetch and validate public URLs. These operations are scoped to user-provided targets and include robust guidelines for concurrency, status code verification, and respecting origin server policies.
[DYNAMIC_EXECUTION]: In the Next.js implementation guide, the skill provides a code snippet for embedding JSON-LD using dangerouslySetInnerHTML. It correctly includes an XSS mitigation pattern (replaceAll('<', '\\u003c') to ensure that user-controlled or external data embedded in the script tag cannot break out of the string context and execute arbitrary code.
[CREDENTIALS_UNSAFE]: The skill explicitly advises against writing credentials or verification tokens into public artifacts and instructs the agent to request scoped access through supported authentication flows, demonstrating a high degree of security awareness.

seo