alterlab-clinicaltrials
Pass
Audited by Gen Agent Trust Hub on Sep 25, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests clinical trial data from the ClinicalTrials.gov API v2. This data includes brief summaries, inclusion/exclusion criteria, and detailed descriptions provided by trial sponsors. If this untrusted external content is directly interpolated into the agent's prompt, it could be used to influence the agent's responses or behavior.
- Ingestion points: Data enters the context via
requests.getcalls inscripts/query_clinicaltrials.py(specifically in thesearch_studiesandget_study_detailsfunctions). - Boundary markers: There are no explicit delimiters or instructions for the agent to ignore potentially malicious content within the fetched trial data.
- Capability inventory: The skill environment permits the use of
Bash(curl:*)andBash(python:*)via its allowed-tools configuration. - Sanitization: The helper script retrieves and returns the raw JSON or text results without performing sanitization or filtering of the natural language fields.
Audit Metadata