osint-methodology
Pass
Audited by Gen Agent Trust Hub on Sep 29, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data (such as web content, JavaScript source, and API responses) to perform reconnaissance and identify vulnerabilities. This creates a potential surface where malicious instructions embedded in these external sources could influence the agent's behavior. * Ingestion points: Data enters the agent's context during the 'Asset Expansion' (Stage 2) and 'Enrichment' (Stage 3) phases as it analyzes external assets. * Boundary markers: No specific delimiters or safety instructions are provided to help the agent ignore instructions found within processed data. * Capability inventory: The skill interacts with companion tools and modules capable of performing network operations and shell execution. * Sanitization: No specific validation or sanitization logic is defined for the external content before it is processed by the agent.
Audit Metadata