perplexity
Pass
Audited by Gen Agent Trust Hub on Oct 5, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill facilitates the ingestion of untrusted data from the open web through Perplexity search tools. This creates a surface for indirect prompt injection where web content could attempt to influence agent behavior.
- Ingestion points:
mcp__perplexity__perplexity_searchandmcp__perplexity__perplexity_askdescribed in SKILL.md. - Boundary markers: The skill does not define specific delimiters for search results in the instructions.
- Capability inventory: The skill is purely instructional and does not perform file writes or system operations itself.
- Sanitization: The skill mandates result limits (max_results: 3) which helps reduce the surface area for external data influence.
Audit Metadata