geo-schema
Pass
Audited by Gen Agent Trust Hub on Oct 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to fetch and analyze external HTML content, creating a surface for indirect prompt injection where malicious instructions could be embedded in the target webpage's source.
- Ingestion points: Target page HTML is fetched using
curlorWebFetchin the detection phase (SKILL.md). - Boundary markers: The instructions lack specific guidance on using delimiters or boundary markers to isolate the external content from the agent's instruction set.
- Capability inventory: The agent identifies, validates, and generates code (JSON-LD) based on the external input. The skill performs network requests to validate discovered URLs (Step 2.6), but does not have dangerous system-level capabilities like file system writes.
- Sanitization: The skill does not define methods for sanitizing or escaping the data extracted from the external HTML before it is processed or reflected in generated reports.
Audit Metadata