seo-geo
Warn
Audited by Gen Agent Trust Hub on Aug 29, 2026
Risk Level: MEDIUMMETADATA_POISONINGINDIRECT_PROMPT_INJECTION
Full Analysis
- [METADATA_POISONING]: The skill and its referenced documents (SKILL.md, references/google-ai-optimization-guide.md) repeatedly use specific dates in the year 2026 (e.g., 'May 2026', 'June 2026', 'Google I/O 2026 keynote') and cite 'future' versions of software (e.g., 'Gemini 2.5'). This provides a deceptive temporal context that may cause the agent to accept simulated or speculative data as historical fact during analysis.
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from user-supplied URLs to perform SEO audits, creating an attack surface for instructions embedded in external content.
- Ingestion points: The [url] argument in SKILL.md is used to fetch and analyze external website content.
- Boundary markers: The instructions lack delimiters or explicit 'ignore embedded instructions' warnings for the data fetched from URLs.
- Capability inventory: The skill uses external tools (DataForSEO MCP) and writes output to GEO-ANALYSIS.md.
- Sanitization: No validation or sanitization of the external content is described.
Audit Metadata