tooluniverse-pharmacovigilance
Pass
Audited by Gen Agent Trust Hub on Mar 8, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill retrieves data from numerous well-known and official scientific repositories, including the FDA (DailyMed, OpenFDA), the NIH (PubMed, ClinicalTrials.gov), and established scientific platforms like ChEMBL, PharmGKB, KEGG, Reactome, and EuropePMC. These sources are documented as trusted scientific domains.\n- [PROMPT_INJECTION]: The skill has a high exposure to potential indirect prompt injection because it ingests and processes large volumes of external text from medical literature and adverse event reports.\n
- Ingestion points: Data is gathered from external sources using tools such as PubMed_search_articles, EuropePMC_search_articles (including preprints), and FAERS_get_event_details.\n
- Boundary markers: The prompt instructions provide structure for the reports but do not include specific delimiters or instructions to treat ingested scientific text as untrusted or to ignore any commands embedded within that content.\n
- Capability inventory: The skill is authorized to write reports in Markdown format and save data to CSV files on the file system, and it can perform subsequent network-based tool calls based on the gathered data.\n
- Sanitization: There is no indication of sanitization, validation, or filtering of the text content retrieved from external databases before it is processed by the agent.\n- [COMMAND_EXECUTION]: The skill uses a comprehensive set of domain-specific tools via the tu.tools framework to execute complex data retrieval and analysis workflows across multiple external platforms.
Audit Metadata