monitor
Pass
Audited by Gen Agent Trust Hub on Feb 17, 2026
Risk Level: SAFEDATA_EXFILTRATION
Full Analysis
- Data Exposure & Exfiltration (LOW): The skill performs network requests to ao-task-monitor.onrender.com, which is a non-whitelisted domain. This is the intended behavior for the CLI tool.
- Indirect Prompt Injection (LOW): The skill processes untrusted external data (API logs and metrics) and provides broad write capabilities via the request command (supporting POST, PUT, DELETE). Evidence Chain: 1. Ingestion: API responses and execution logs. 2. Boundary markers: Absent. 3. Capability inventory: Arbitrary HTTP requests via the request command. 4. Sanitization: Not specified in documentation.
- Persistence Mechanisms (LOW): The documentation instructs users to manually modify shell configuration files (.zshrc, .bashrc) to persist API keys, which is a common but sensitive system configuration task.
Audit Metadata