tooluniverse-cell-line-profiling
Pass
Audited by Gen Agent Trust Hub on Oct 7, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONDYNAMIC_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes data from a variety of external biological databases. This creates a surface where malicious instructions hidden within scientific records could influence the agent's logic.\n
- Ingestion points: Data retrieved from ToolUniverse tools (e.g., COSMIC, PharmacoDB) and large CSV datasets downloaded from DepMap.\n
- Boundary markers: The skill does not define delimiters or specific instructions to separate untrusted external data from its own processing logic.\n
- Capability inventory: The skill allows the agent to generate and execute Python code via a Bash environment for statistical analysis.\n
- Sanitization: No specific data sanitization or schema validation processes are mentioned for the external API outputs.\n- [EXTERNAL_DOWNLOADS]: The
scripts/depmap_gene_dependency.pyscript downloads genomic dataset files from the DepMap portal (depmap.org), which is a well-known scientific resource.\n - Evidence: The script uses
urllib.request.urlretrieveto fetch data from the portal's download index.\n- [COMMAND_EXECUTION]: The skill instructions direct the agent to use shell commands for data analysis tasks.\n - Evidence: The prompt explicitly tells the agent to 'write and run Python code via Bash' to execute computation rather than just describing it.\n- [DYNAMIC_EXECUTION]: The skill facilitates the runtime generation of Python scripts to handle complex data analysis needs.\n
- Evidence: The 'COMPUTE, DON'T DESCRIBE' section instructs the agent to write and execute code dynamically to report actual results.
Audit Metadata