sn-deep-research
Fail
Audited by Gen Agent Trust Hub on Nov 3, 2026
Risk Level: CRITICALCOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill orchestrates a complex multi-agent workflow using shell commands, Python scripts, and a Node.js workbench runtime to create report directories and manage research progress.
- [EXTERNAL_DOWNLOADS]: The research agent is designed to scrape and download content from a wide array of external sources including academic repositories, financial filings, and social media platforms to aggregate evidence for its reports.
- [INDIRECT_PROMPT_INJECTION]: Because the skill processes untrusted web data, it has a significant attack surface for indirect prompt injection. However, it mitigates this by including explicit safety instructions in the 'review' agent's prompt to ignore instructions found in external content and treat them only as data for verification.
- [SAFE]: The skill follows security best practices by instructing users to manage credentials via environment variables and .env files, using structured JSON schemas with automated validation to prevent malformed data propagation, and maintaining clear boundaries between untrusted data and agent instructions.
Recommendations
- Contains 1 malicious URL(s) - DO NOT USE
Audit Metadata