alterlab-bindingdb
Pass
Audited by Gen Agent Trust Hub on Sep 25, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill interacts with legitimate, well-known scientific databases including BindingDB and PubChem to retrieve experimental binding affinities. The code snippets and scripts use standard Python libraries like requests and pandas for data retrieval and processing, following established best practices.
- [INDIRECT_PROMPT_INJECTION]: The skill processes data from external sources (BindingDB REST API and TSV files). While it lacks explicit prompt boundary markers for all ingested data, the source URLs are reputable scientific repositories, and the data is handled using structured methods (JSON parsing and pandas dataframes), which significantly limits the risk of prompt-based attacks.
Audit Metadata