cardiology-editorial
Pass
Audited by Gen Agent Trust Hub on Nov 6, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted clinical data from external sources, which is a primary surface for indirect instructions to influence agent behavior.
- Ingestion points: The workflow relies on clinical trial abstracts fetched via
PubMed:search_articlesand full-text manuscripts provided through user-uploaded PDFs. - Boundary markers: There are no explicit delimiters (such as XML tags or triple backticks) or instructions to disregard embedded commands when interpolating retrieved trial data into the LLM prompts for classification and importance scoring.
- Capability inventory: The skill uses
PubMedMCP tools to search and retrieve metadata, and it generates 500-word authoritative editorial content based on that external information. - Sanitization: The instructions do not define any sanitization or validation steps for the content of retrieved abstracts or uploaded files before they are processed by the scoring and writing modules.
Audit Metadata