tooluniverse-target-research
Pass
Audited by Gen Agent Trust Hub on Jul 30, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [SAFE]: The skill retrieves biological and chemical information from established and reputable organizations such as UniProt, Ensembl, Pharos, and Open Targets. These operations represent standard research functionality and follow expected patterns for bioinformatics tools.
- [PROMPT_INJECTION]: The skill possesses a surface for indirect prompt injection because its core function involves ingesting and summarizing extensive external data from diverse sources such as PubMed and Open Targets.
- Ingestion points: The agent ingests data from numerous scientific API tools across 9 research paths, specifically fetching literature and database annotations through tools like
PubMed_search_articlesandOpenTargets_get_publications_by_target_ensemblId(SKILL.md, Path 0 and Path 8). - Boundary markers: The skill requires the agent to follow a strict markdown report format (
[TARGET]_target_report.md) with predefined headers, which acts as a structural boundary, but the instructions lack explicit directives to sanitize data or ignore potential command instructions embedded within scientific text (REPORT_FORMAT.md). - Capability inventory: The skill empowers the agent to write complex reports to the file system and execute a wide array of specialized analytical tools (SKILL.md, Phase 1-8).
- Sanitization: No specific input sanitization, escaping mechanisms, or filtering logic is prescribed for processing the external data before it is displayed to the user or saved in the final intelligence report.
Audit Metadata