info-search-knowledge
Pass
Audited by Gen Agent Trust Hub on Feb 17, 2026
Risk Level: SAFE
Full Analysis
- Indirect Prompt Injection (LOW): The skill defines a workflow for ingesting untrusted data from external websites (Official blogs, Reddit, GitHub, tech media). This creates an attack surface where malicious content on a searched page could attempt to influence the agent's behavior. Evidence: Step 3 (Search Execution) describes systematic data collection. Mitigation: Step 4 (Quality Verification) includes source and content validation, serving as a procedural defense.
- No Code (SAFE): The skill consists entirely of Markdown instructions. No scripts (Python, Node.js), binaries, or shell commands are included or referenced for execution.
- Data Exfiltration (SAFE): While the skill describes saving data to a knowledge base, it provides no mechanisms for outbound network requests to untrusted domains or unauthorized access to sensitive local files.
Audit Metadata