The Agent Skills Directory

[Indirect Prompt Injection] (LOW): The script scripts/prepare_batch_csv.py processes external CSV files provided by users, which is a potential entry point for malicious instructions aimed at the agent processing the data.
Ingestion points: The pd.read_csv(csv_path) call in scripts/prepare_batch_csv.py accepts external data files.
Boundary markers: None identified; the skill does not use specific delimiters or instructions to prevent the agent from following directions embedded in the CSV content.
Capability inventory: No dangerous capabilities like arbitrary command execution (eval, subprocess) or unauthorized network requests were detected in the provided scripts.
Sanitization: The script performs validation on chemical structure strings (SMILES) and verifies file paths, but it does not include sanitization against prompt-based attacks.
[External Downloads] (SAFE): Documentation and environment scripts reference official scientific software repositories on GitHub and standard registries like PyPI. No automated downloads of untrusted scripts were found.
[Data Exposure] (SAFE): No credentials, private keys, or sensitive local configuration paths (e.g., .aws/credentials) were detected in the skill's source files.
[Obfuscation] (SAFE): No obfuscated code, zero-width characters, or multi-layer encoding was found.

diffdock