tooluniverse-chemical-safety
Pass
Audited by Gen Agent Trust Hub on Oct 7, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to write and execute Python code via a Bash interface to perform data processing, statistical analysis, and visualization. This capability is used to generate reports from chemical data and is part of the core functionality.
- [EXTERNAL_DOWNLOADS]: The documentation references the installation of the
tooluniverse[ml]Python package. This is a platform-associated dependency required for the machine learning features of the ADMET-AI tools. - [DYNAMIC_EXECUTION]: The skill employs runtime generation of Python scripts to analyze data retrieved from toxicology databases, which is the intended mechanism for creating detailed safety dossiers.
- [INDIRECT_PROMPT_INJECTION]: The skill has an attack surface for indirect prompt injection as it ingests untrusted chemical names and SMILES strings from user queries.
- Ingestion points: Chemical identifiers and SMILES strings in
SKILL.mdresearch triggers. - Boundary markers: Absent for raw chemical strings.
- Capability inventory: Shell execution (Bash) for Python scripts and access to over 25 bioinformatics tools.
- Sanitization: The skill includes explicit instructions to verify tool parameters using metadata (
get_tool_info) before execution, providing a layer of validation against malformed inputs.
Audit Metadata