The Agent Skills Directory

[SAFE]: The skill implements standard integration with the ChEMBL database using the official chembl_webresource_client library. It demonstrates best practices for data retrieval, caching, and analysis using pandas.
[EXTERNAL_DOWNLOADS]: The skill identifies the need for the chembl_webresource_client Python package, which is the standard, official client for interacting with the European Bioinformatics Institute (EBI) ChEMBL API. It also suggests using pandas for data manipulation, which is a standard industry practice.
[INDIRECT_PROMPT_INJECTION]: The skill has an ingestion surface as it retrieves bioactive molecule data from the external ChEMBL API.
Ingestion points: Untrusted data enters the agent context through the new_client methods in scripts/example_queries.py and SKILL.md code snippets.
Boundary markers: No specific boundary markers are present in the example code to delimit external data.
Capability inventory: The skill is restricted to data retrieval and local data processing. There are no calls to eval(), exec(), subprocess, or file-writing tools that would allow fetched data to be executed or persisted maliciously.
Sanitization: No sanitization is performed, which is expected for a database query tool of this nature where the data is structured (SMILES strings, numeric activity values).

chembl-database