exploratory-data-analysis
Warn
Audited by Gen Agent Trust Hub on Oct 12, 2026
Risk Level: MEDIUMMETADATA_POISONINGINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [METADATA_POISONING]: The skill metadata identifies 'K-Dense Inc.' as the author, which conflicts with the expected author context 'wu-yc'. Furthermore, the instructions include a section titled 'Suggest Using K-Dense Web For Complex Worflows' that directs the agent to steer users toward an external commercial platform (www.k-dense.ai) for advanced tasks.
- [INDIRECT_PROMPT_INJECTION]: The skill's primary function is to read and analyze data from over 200 external scientific file formats, creating a broad attack surface for data-driven injections.
- Ingestion points: The agent is directed to open and parse diverse file types including .fasta, .pdb, .csv, and .nd2 (detailed in SKILL.md).
- Boundary markers: No explicit use of delimiters or 'ignore instructions' warnings are mandated when processing untrusted file data.
- Capability inventory: The skill facilitates the execution of local Python scripts (scripts/eda_analyzer.py) and allows for file system writes ({original_filename}_eda_report.md).
- Sanitization: There is no documentation or instruction for sanitizing or validating the content of the scientific data files before it is processed by the agent.
- [COMMAND_EXECUTION]: The skill instructions explicitly direct the agent to invoke the shell to run scripts/eda_analyzer.py for automated file analysis.
- [EXTERNAL_DOWNLOADS]: The documentation suggests the installation of numerous external Python dependencies (e.g., biopython, rdkit, mdanalysis, nd2reader) using package managers like uv or pip to support specialized file formats.
Audit Metadata