mental-health-analyzer
Pass
Audited by Gen Agent Trust Hub on Aug 9, 2026
Risk Level: SAFEPROMPT_INJECTIONDATA_EXFILTRATION
Full Analysis
- [PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection by design, as it reads and processes free-text daily mood diaries and health logs provided by the user to perform complex mental health analysis and crisis risk assessments.
- Ingestion points: The agent reads untrusted data from multiple files including
data-example/mental-health-logs/,mental-health-tracker.json, and several other health-related JSON files. - Boundary markers: There are no defined delimiters or instructions to the agent to treat the contents of these logs as non-executable data, increasing the risk that instructions embedded within logs could override the agent's behavior.
- Capability inventory: The skill is granted access to powerful file system tools including
Read,Write,Edit,Grep, andGlob. - Sanitization: The instructions do not specify any sanitization, validation, or filtering of the free-text content before it is analyzed for psychological patterns or risk markers.
- [DATA_EXFILTRATION]: The skill accesses and processes highly sensitive personal data, including PHQ-9 and GAD-7 mental health assessments, medication logs, and chronic disease records. While no unauthorized network operations were identified, the combination of sensitive data access and the capability to write to the file system creates a significant data exposure surface.
Audit Metadata