purchase-research

Pass

Audited by Gen Agent Trust Hub on Sep 3, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from external sources, creating a potential surface for indirect prompt injection.
  • Ingestion points: The skill retrieves data using search_policies, search_products, and web_search tools as described in SKILL.md.
  • Boundary markers: There are no explicit instructions or delimiters configured to prevent the agent from potentially obeying instructions embedded within the retrieved external content.
  • Capability inventory: The skill has access to presentation tools (present_guide, present_comparison, present_products) and retrieval tools (web_search, get_product_details). It does not have access to high-risk tools like shell execution or local file-system modification.
  • Sanitization: The instructions mandate that web material be "restated in your own words," which provides a natural language processing layer but does not constitute a formal security boundary against adversarial content.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 3, 2026, 05:23 PM
Security Audit — agent-trust-hub — purchase-research