anysite-competitor-intelligence
Pass
Audited by Gen Agent Trust Hub on Feb 16, 2026
Risk Level: LOWPROMPT_INJECTION
Full Analysis
- [Indirect Prompt Injection] (LOW): The file describes patterns for ingesting untrusted external data. 1. Ingestion points: Reddit and LinkedIn posts via anysite MCP tools. 2. Boundary markers: Absent in the documentation. 3. Capability inventory: Focused on data retrieval and reasoning; no write or execute capabilities found. 4. Sanitization: Absent. This surface allows external content to influence agent reasoning.
- [Data Exposure] (SAFE): No access to sensitive local files or hardcoded credentials detected.
- [Remote Code Execution] (SAFE): No remote code execution or dynamic code generation patterns identified.
Audit Metadata