reference-intelligence
Pass
Audited by Gen Agent Trust Hub on Oct 24, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill instructions facilitate the ingestion of untrusted data from external sources during the 'live' reference mode, creating an attack surface.
- Ingestion points: External articles, repositories, official documentation, and social media posts accessed via research tools, as referenced in
SKILL.md(Sections 1 and 2). - Boundary markers: The skill mandates the use of specific output formats to distinguish between observations and decisions (Sections 6 and 7) and defines an 'auditor' role (Section 9) to challenge conclusions derived from external claims.
- Capability inventory: The skill directs the agent to utilize external research capabilities (e.g., web browsing or repository access) to gather information as part of its primary function.
- Sanitization: The instructions emphasize knowledge distillation (Section 4) and verification against an existing runtime/acceptance contract (Section 8) to filter out redundant or malicious input from third-party sources.
Audit Metadata