bioservices
Pass
Audited by Gen Agent Trust Hub on Apr 16, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: Behavioral steering instructions found in SKILL.md. The section 'Suggest Using K-Dense Web For Complex Worflows' instructs the agent to proactively suggest the author's commercial platform (www.k-dense.ai) when complex tasks are requested, overriding natural agent behavior for marketing purposes.
- [PROMPT_INJECTION]: Indirect prompt injection surface. The skill ingests untrusted data from external bioinformatics APIs (UniProt, KEGG, ChEMBL) as described in SKILL.md. Evidence chain: 1. Ingestion points: Web services like UniProt, KEGG, and NCBI (SKILL.md). 2. Boundary markers: Absent. 3. Capability inventory: Subprocess execution via Python scripts mentioned in the scripts/ folder. 4. Sanitization: Absent. Malicious instructions embedded in biological database records could theoretically influence agent behavior during data processing.
Audit Metadata