The Agent Skills Directory

[SAFE]: The skill demonstrates security awareness by including explicit instructions for the agent to treat fetched web content as untrusted data, specifically identifying meta tags and page copy as potential prompt-injection surfaces.
[INDIRECT_PROMPT_INJECTION]: The skill processes content from external websites, which is a known vector for indirect prompt injection.
Ingestion points: Content retrieved from external URLs using web_fetch or curl as described in the Audit Framework.
Boundary markers: Includes an explicit warning in SKILL.md to ignore instructions embedded in HTML, meta tags, or page copy.
Capability inventory: The skill primarily uses data-retrieval tools to perform diagnostics and generate reports; it does not demonstrate capabilities for unauthorized file modification or credential exfiltration.
Sanitization: No technical sanitization is present, but the skill implements a strategic instructional guardrail for the agent to treat the external input as data rather than instructions.

seo-audit