cosmic-database
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests and processes external genomic data from the COSMIC database, which constitutes a surface for indirect prompt injection. While the data is scientific and structured (TSV, VCF), the skill lacks explicit sanitization or boundary markers for content that might be processed by the agent. This risk is considered negligible given the biological research context and structured nature of the files.
- Ingestion points: Cancer mutation files and gene census lists downloaded via
scripts/download_cosmic.py. - Boundary markers: No delimiters are specified in the provided data analysis examples.
- Capability inventory: The skill performs network requests for data retrieval and writes files to the local system.
- Sanitization: No explicit content filtering is applied to the biological datasets before analysis.
- [SAFE]: The skill communicates with the official and well-known scientific domain
cancer.sanger.ac.uk. This network activity is the primary intended function for data acquisition and is conducted using standard authenticated requests.
Audit Metadata