The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes external molecular data which creates a surface for indirect prompt injection. 1. Ingestion points: Molecular structures and metadata are loaded from external SDF, SMILES, and MOL files using RDKit suppliers in SKILL.md and the provided utility scripts. 2. Boundary markers: There are no specific delimiters or instructions for the agent to ignore potentially malicious textual instructions embedded in the metadata fields of these chemical data files. 3. Capability inventory: The skill is configured with access to the Bash tool and file operations (Read, Write, Edit), providing a significant capability set if an injection were successful. 4. Sanitization: While the skill utilizes RDKit's chemical sanitization (Chem.SanitizeMol) to ensure molecular validity, it does not perform security-focused sanitization of textual metadata fields.
[COMMAND_EXECUTION]: The skill's configuration in SKILL.md allows the use of the Bash tool. This is used for standard installation procedures (e.g., uv pip install rdkit) and for running the included Python analysis scripts, but it grants the agent the ability to execute arbitrary shell commands in the execution environment.

rdkit