The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill's documentation and examples in SKILL.md and references/examples.md demonstrate caching molecular embeddings using the Python pickle module. The use of pickle.load() to deserialize data is a known security risk that can lead to arbitrary code execution if the local cache file is replaced with a malicious payload by an attacker.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process molecular data (SMILES strings) from external datasets, which constitutes an attack surface for indirect prompt injection.
Ingestion points: Data enters the agent context through the MoleculeTransformer and various calculator classes in molfeat.calc that accept lists of SMILES strings or RDKit molecule objects.
Boundary markers: The provided examples do not implement delimiters or explicit instructions to the AI to ignore potential commands embedded in the processed molecular data.
Capability inventory: The skill possesses the capability to write configuration files to the file system (to_state_yaml_file, to_state_json_file) and perform file operations for caching embeddings.
Sanitization: There is no evidence of input validation or sanitization for the SMILES strings before they are processed by the featurization calculators.

molfeat