crewai-agents
Pass
Audited by Gen Agent Trust Hub on Jun 26, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [COMMAND_EXECUTION]: Includes a code example for a
DatabaseQueryToolthat demonstrates how to execute SQL queries usingpandas.read_sql. This is presented as an educational pattern for extending agent capabilities.\n- [EXTERNAL_DOWNLOADS]: References the use ofSerperDevTool, which is a standard utility for performing web searches via external APIs.\n- [PROMPT_INJECTION]: The skill documents a multi-agent pipeline (Researcher, Writer, Reviewer) where data retrieved from external sources is processed sequentially, creating a surface for indirect instructions.\n - Ingestion points:
SerperDevTool(web results) andFileReadTool(local file system) inSKILL.md.\n - Boundary markers: No explicit instruction-data delimiters are shown in the task templates.\n
- Capability inventory: File system access via
FileReadTooland SQL execution via the custom database tool example.\n - Sanitization: No explicit sanitization or validation of data retrieved from external tools is demonstrated in the examples.
Audit Metadata