bio-data-visualization-manhattan-qq-locuszoom
Pass
Audited by Gen Agent Trust Hub on Oct 8, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSDATA_EXFILTRATIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill provides instructions to download and install standard scientific packages from official repositories, including CRAN and PyPI.
- [DATA_EXFILTRATION]: The skill interacts with the external LDlink API using the
LDlinkRlibrary. It implements best practices by advising the use of environment variables for API token management. - [INDIRECT_PROMPT_INJECTION]: The skill ingests and processes user-supplied GWAS summary statistics files, creating a vector for potential indirect prompt injection attacks.
- Ingestion points: Data is read from local summary statistics files (e.g.,
gwas_summary.tsv) using standard R functions. - Boundary markers: No specific boundary markers or 'ignore' instructions are utilized when importing external data.
- Capability inventory: The agent can read local files, write resulting plot files, and perform network requests via the LDlink service.
- Sanitization: The skill performs standard data manipulation without specific sanitization of the input text for instruction-like patterns.
Audit Metadata