skills/freedomintelligence/openclaw-medical-skills/tooluniverse-clinical-guidelines/Gen Agent Trust Hub
tooluniverse-clinical-guidelines
Pass
Audited by Gen Agent Trust Hub on Oct 7, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process clinical guidelines from over 12 external authoritative sources, including NICE, WHO, and PubMed. This architecture creates an indirect prompt injection surface where the agent could potentially be influenced by malicious instructions embedded in external medical content.
- Ingestion points: Multiple tools fetch data from external APIs (e.g.,
GIN_Guidelines_Search,TRIP_Database_Guidelines_Search,PubMed_Guidelines_Search) as defined inSKILL.md. - Boundary markers: The skill instructions do not provide specific delimiters or "ignore" instructions to prevent the agent from obeying commands found within the retrieved medical text.
- Capability inventory: The skill possesses extensive data retrieval and synthesis capabilities, including tools for full-text extraction (
NICE_Guideline_Full_Text,AHA_ACC_get_guideline,NCCN_get_patient_guideline). - Sanitization: No sanitization or filtering of external content is specified before interpolation into the agent's response.
- [DYNAMIC_EXECUTION]: The "Phase 0: Tool Verification" section in
SKILL.mdinstructs the agent to execute Python code at runtime to verify the loading of thetooluniverselibrary. While this is a common pattern for confirming tool availability, it involves the dynamic generation and execution of code within the agent's environment.
Audit Metadata