alterlab-rdkit
Warn
Audited by Gen Agent Trust Hub on Apr 12, 2026
Risk Level: MEDIUMCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The 'Best Practices' section of 'SKILL.md' suggests using the 'pickle' module for serializing and deserializing molecular data for performance. The 'pickle.load()' function is inherently insecure and can result in arbitrary code execution if used on files provided by an untrusted source.
- [PROMPT_INJECTION]: The skill provides scripts for processing external data (SMILES, SDF, MOL), which serves as a surface for indirect prompt injection. Ingestion points: The scripts 'molecular_properties.py', 'similarity_search.py', and 'substructure_filter.py' ingest molecule data via RDKit supplier classes. Boundary markers: There are no markers used to isolate data from potential instructions. Capability inventory: The skill possesses the ability to perform file system writes and complex data operations. Sanitization: Input is validated for chemical validity but not for security-threatening natural language patterns.
Audit Metadata