The Agent Skills Directory

[SAFE]: The skill is designed to perform SEO audits which necessitates fetching and processing untrusted content from external websites. This activity creates an inherent surface for indirect prompt injection, which is a standard risk for auditing and summarization tasks. Analysis of the risk factors shows:
Ingestion points: External site content is retrieved via web_fetch or curl instructions.
Boundary markers: No specific delimiters or "ignore instructions" wrappers are implemented for the fetched content.
Capability inventory: The skill identifies capabilities for local file reading (.agents/product-marketing-context.md) and network operations (curl, web_fetch).
Sanitization: No explicit sanitization or validation of the fetched HTML/metadata is defined. Despite this surface, the behavior is strictly aligned with the skill's stated purpose and presents no malicious patterns.
[SAFE]: The skill correctly identifies and warns the agent about technical limitations, such as the inability of static fetch tools to detect JavaScript-injected schema markup, and recommends using well-known, trusted services like the Google Rich Results Test for accurate validation.
[SAFE]: All referenced external documentation and tools (Google Search Central, Search Engine Journal, Semrush, etc.) are from trusted organizations or well-known services in the technology and SEO industries.
[SAFE]: The skill follows secure practices by encouraging the use of standard project context files for personalization rather than requesting sensitive credentials or system-level access.

seo-audit