perspective-taking-designer
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill incorporates several user-controlled text inputs directly into the primary instruction prompt used to generate activities.\n- Ingestion points: Untrusted data enters the agent context through the
subject_context,learning_goal,age_range,sensitivity_notes, andprior_knowledgefields defined in SKILL.md.\n- Boundary markers: The prompt template does not utilize delimiters, XML-style tags, or explicit instructions to ignore or isolate potential commands embedded within the user-provided inputs.\n- Capability inventory: The skill contains no instructions or configuration for network requests, file system access, subprocess spawning, or other dangerous operations across all sections.\n- Sanitization: There is no evidence of input validation, filtering, or escaping logic applied to the data received from the user before it is processed by the model.
Audit Metadata