The Agent Skills Directory

Dynamic Execution (MEDIUM): The skill utilizes the 'pickle' module to cache parsed DrugBank data in references/data-access.md.
Evidence: The file references/data-access.md includes code snippets using pickle.load(f) on a local file.
Risk: Deserializing data using 'pickle' can lead to arbitrary code execution if the cache file is tampered with by a malicious actor.
Indirect Prompt Injection (LOW): The skill is designed to ingest and parse large XML datasets from external sources.
Ingestion points: Data is retrieved via drugbank-downloader and parsed in references/data-access.md.
Boundary markers: No explicit boundary markers or 'ignore' instructions are provided for the ingested XML content.
Capability inventory: The skill has network access (requests), file system access (writing cache files), and XML parsing capabilities.
Sanitization: No evidence of sanitization or safety filtering for the content of the drug database before it is processed by the agent.
External Downloads (SAFE): The skill downloads data from go.drugbank.com.
Evidence: Referenced in SKILL.md and references/data-access.md.
Risk: While the source is reputable, the skill relies on external data that is not under the immediate control of the user.

drugbank-database