The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill uses the 'pickle' module in its data caching implementation. The use of 'pickle.load()' to deserialize local data is a known security risk as it can facilitate arbitrary code execution if a user or attacker replaces the cached file with a malicious payload. Evidence found in 'references/data-access.md'.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process external drug information from the DrugBank database. This architecture creates an indirect prompt injection surface where malicious data within the database could potentially influence the agent's behavior.
Ingestion points: The skill fetches XML data from 'go.drugbank.com' using the 'drugbank-downloader' package.
Boundary markers: No explicit delimiters or instructions to ignore embedded commands are present in the scripts or documentation.
Capability inventory: The skill possesses file-writing capabilities through 'pandas.to_csv' and network access via the 'requests' library.
Sanitization: There is no evidence of sanitization or safety checks performed on the data fields extracted from the XML.
[CREDENTIALS_UNSAFE]: The skill provides instructions for managing DrugBank user credentials. While it suggests standard practices like environment variables, it also references the use of a local configuration file at '~/.config/drugbank.ini', which is a sensitive file path containing authentication secrets as seen in 'references/data-access.md'.

drugbank-database