disease-drug-intelligence

Pass

Audited by Gen Agent Trust Hub on Mar 25, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill makes network requests to well-known and reputable scientific data providers: www.ebi.ac.uk (ChEMBL API) and clinicaltrials.gov. It also utilizes the Tavily Search API for supplemental information gathering.\n- [COMMAND_EXECUTION]: The skill includes a shell script local_tools/run_tool.sh used to execute local Python scripts (chembl_api.py, clinicaltrials_api.py, search_api.py) for data processing and API interaction.\n- [PROMPT_INJECTION]: The skill contains a surface for indirect prompt injection as it ingests and processes content from external web searches and pharmaceutical databases. \n
  • Ingestion points: Data from external APIs is pulled into the agent's context via local_tools/chembl_api.py, local_tools/clinicaltrials_api.py, and local_tools/search_api.py.\n
  • Boundary markers: The skill instructions in SKILL.md explicitly mandate labeling evidence boundaries ("标注证据边界") to distinguish external data from internal reasoning.\n
  • Capability inventory: The skill can execute local shell commands via run_tool.sh and perform network operations via Python's requests library.\n
  • Sanitization: No explicit programmatic sanitization or filtering of API responses is implemented in the provided Python source code.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 25, 2026, 05:57 AM