The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests molecular data from external files and remote URLs (S3, GCS, HTTP), which creates a potential surface for indirect prompt injection if metadata fields within these files contain malicious instructions.\n
Ingestion points: dm.read_sdf, dm.read_csv, dm.read_excel, and dm.open_df as documented in SKILL.md and references/io_module.md.\n
Boundary markers: No explicit boundary markers or instructions to ignore embedded commands are present in the documentation.\n
Capability inventory: The skill has file-writing capabilities (dm.to_sdf, dm.save_df), network access via fsspec protocols, and parallel processing support.\n
Sanitization: While the skill emphasizes molecular standardization (dm.standardize_mol), it does not mention sanitization for textual metadata fields.\n- [EXTERNAL_DOWNLOADS]: The skill instructions direct the user to install the datamol package from public registries.

datamol