start
Pass
Audited by Gen Agent Trust Hub on Feb 19, 2026
Risk Level: SAFE
Full Analysis
- [Prompt Injection] (SAFE): The instructions are focused on guiding a specific diagnostic conversation. No override, bypass, or role-play injection patterns were found.
- [Data Exposure & Exfiltration] (SAFE): The skill directs the agent to save data to
data/learner-profile.json. It specifically avoids requesting personally identifiable information (PII) such as names or contact details. There are no network operations or sensitive file path accesses. - [Indirect Prompt Injection] (LOW): The skill ingests untrusted user input (survey answers) and stores them in a JSON file. While this data is later used for learning path recommendations, the interaction is confined to a structured 5-question format, minimizing the surface for prompt-based manipulation of agent behavior.
- [No Code] (SAFE): No external dependencies, scripts, or executable code are included or referenced in the skill definition.
Audit Metadata