research
Pass
Audited by Gen Agent Trust Hub on Oct 14, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDATA_EXFILTRATION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data retrieved via web searches, which constitutes an ingestion point for potentially malicious instructions. However, it incorporates a robust guardrail explicitly instructing the agent to treat all scraped text as actively hostile and to disregard any commands or instructions found within research context.
- [DATA_EXFILTRATION]: The skill facilitates network operations through search engines (Brave/Tavily). It includes a mandatory step to sanitize search metadata, ensuring that internal project codenames, API keys, or structural details are stripped before the query is sent externally.
- [SAFE]: No malicious patterns such as obfuscation, hardcoded credentials, unauthorized command execution, or privilege escalation were detected in the skill definition or its documentation.
Audit Metadata