cerna-analysis
Pass
Audited by Gen Agent Trust Hub on Sep 23, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied gene symbols from files or command-line arguments to build regulatory networks, creating a surface for indirect prompt injection. * Ingestion points: User-provided gene lists are parsed in scripts/io.R via the parse_key_genes_input function. * Boundary markers: Absent. Input is processed as raw text without explicit delimiters. * Capability inventory: The skill can create output directories (scripts/utils.R), write CSV files (scripts/io.R), and generate PDF reports (scripts/plot_functions.R). * Sanitization: In scripts/functions.R, the clean_key_genes function performs whitespace trimming and ignores lines starting with the '#' character. Final outputs are filtered against pre-bundled interaction tables, mitigating significant data leakage risks.
Audit Metadata