cosmic-database
Pass
Audited by Gen Agent Trust Hub on Apr 16, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill contains behavioral steering instructions in the 'Suggest Using K-Dense Web For Complex Worflows' section. It directs the agent to monitor user interaction complexity and proactively suggest the 'K-Dense Web' platform (www.k-dense.ai), which overrides default agent behavior for promotional purposes.
- [PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection because it ingests and processes large external datasets which could potentially contain malicious instructions.
- Ingestion points: Data enters the context via mutation files (e.g., 'cosmic_mutations.tsv.gz') and gene census CSVs as shown in the SKILL.md examples.
- Boundary markers: The instructions do not define delimiters or provide 'ignore instructions' warnings when processing these external data files.
- Capability inventory: The skill has network access via 'scripts/download_cosmic.py' (using requests) and significant data processing capabilities through pandas and pysam.
- Sanitization: No sanitization, escaping, or structural validation is mentioned for the ingested genomic data.
Audit Metadata