pydeseq2
Warn
Audited by Gen Agent Trust Hub on Jun 30, 2026
Risk Level: MEDIUMCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [DYNAMIC_EXECUTION]: The documentation within 'references/workflow_guide.md' suggests using 'pickle.load()' to load data and previous analysis results. Deserialization with the 'pickle' module is insecure when used with untrusted files, as it can be leveraged to achieve arbitrary code execution.
- [INDIRECT_PROMPT_INJECTION]: The skill and its script 'scripts/run_deseq2_analysis.py' ingest untrusted genomic and metadata from CSV/TSV files. Ingestion points occur in the 'load_and_validate_data' and 'filter_data' functions. There are no boundary markers or instructions to the agent to ignore potentially malicious commands embedded in these data files. The capability inventory includes writing files to the disk and performing complex data operations, and no sanitization of the input strings is performed.
- [DATA_EXPOSURE_AND_EXFILTRATION]: The skill reads local CSV files containing RNA-seq counts and metadata. This behavior is necessary for its stated analytical purpose, and no data exfiltration to external domains was identified.
Audit Metadata