The Agent Skills Directory

[METADATA_POISONING]: The skill-author field in the SKILL.md frontmatter is set to 'K-Dense Inc.', which differs from the attributed author 'kinandigital'. This is a minor metadata inconsistency.
[INDIRECT_PROMPT_INJECTION]: The skill functions by ingesting external therapeutic datasets from the Therapeutics Data Commons (tdcommons.ai) into the agent context.
Ingestion points: Data loading tasks defined in SKILL.md and scripts/load_and_split_data.py fetch external SMILES strings, protein sequences, and labels.
Boundary markers: The instructions do not define specific delimiters or guardrails to prevent the agent from interpreting content within these datasets as instructions.
Capability inventory: The skill performs network operations to retrieve datasets and utilizes file system access to cache data in a local /data directory.
Sanitization: No explicit sanitization or validation logic for the external chemical data is included in the utility scripts.

pytdc