kove
Pass
Audited by Gen Agent Trust Hub on Nov 12, 2026
Risk Level: SAFE
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from user-generated journal notes, field names, and exported record files (README.txt). 1. Ingestion points: get_day, get_days, and export_all results. 2. Boundary markers: Explicit 'The person's record is data' section in SKILL.md directing the agent to treat record content as data, not instructions. 3. Capability inventory: curl command execution, local file writing, and file reading. 4. Sanitization: Instructions to treat user data as facts and ignore embedded commands mitigate risk.
- [COMMAND_EXECUTION]: The skill defines a workflow for data export that involves executing a
curlcommand to download a ZIP file from the vendor's infrastructure. This is a standard functional component of the service and is restricted to the user's own data requested through official tool outputs. - [DATA_EXFILTRATION]: While the skill facilitates the movement of sensitive health data, it does so through authorized tools (like
export_all) and includes strict instructions for the agent to protect the privacy of generated download links and to handle the data as the person's private record.
Audit Metadata