tooluniverse-admet-prediction

Pass

Audited by Gen Agent Trust Hub on Sep 16, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user inputs (drug names and SMILES strings) which are subsequently passed to external bioinformatics tools and potentially interpolated into dynamically generated analysis scripts.
  • Ingestion points: User-provided 'Drug name' or 'SMILES string' defined in SKILL.md.
  • Boundary markers: Absent. There are no explicit instructions to treat the inputs as raw data or ignore embedded instructions.
  • Capability inventory: Python execution via Bash, network access to PubChem, SwissADME, and ChEMBL APIs.
  • Sanitization: Absent. The skill does not define validation steps for the input strings beyond basic resolution.
  • [DYNAMIC_EXECUTION]: The skill explicitly instructs the agent to "write and run Python code via Bash" for data processing, statistics, and enrichment tasks using libraries such as pandas and scipy.
  • This constitutes runtime generation and execution of code based on retrieved data.
  • [COMMAND_EXECUTION]: The agent is directed to use the system shell (Bash) to execute Python scripts and diagnostic tools like tooluniverse-doctor.
  • [EXTERNAL_DOWNLOADS]: The documentation includes setup instructions to install the tooluniverse[ml] and admet-ai packages from standard package registries. These are necessary components for the skill's machine-learning-based predictions.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 16, 2026, 06:31 AM
Security Audit — agent-trust-hub — tooluniverse-admet-prediction