The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill documentation provides instructions for installing standard scientific Python packages from PyPI, including datamol, s3fs, gcsfs, scipy, and scikit-learn. These are well-known libraries in the cheminformatics and data science communities.
[DATA_EXFILTRATION]: The skill supports reading from and writing to remote storage protocols (S3, GCS, Azure, and HTTPS). While this creates a data transfer capability, the instructions include proactive security guidance, such as requiring user confirmation for remote write paths and ensuring credentials are used locally through the standard provider environment variables.
[COMMAND_EXECUTION]: The skill uses the Bash tool to perform initial dependency installation using the uv package manager as described in SKILL.md.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process external molecular data files (such as SDF, CSV, and SMILES) from local or remote URLs. This capability introduces a surface for indirect prompt injection if external files contain malicious instructions.
Ingestion points: Functions like dm.read_sdf(), dm.read_csv(), and dm.open_df() in SKILL.md and references/io_module.md that accept remote URLs.
Boundary markers: The skill documentation focuses on chemical structure validation rather than natural language delimiters for data content.
Capability inventory: The agent has capabilities for writing files (SKILL.md), network access for cloud storage (references/io_module.md), and executing shell commands for setup (SKILL.md).
Sanitization: Recommends dm.standardize_mol() and dm.sanitize_mol() to validate chemical data, but these do not filter for injection patterns in text fields.

datamol