The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes a wide variety of scientific file formats (e.g., CSV, FASTA, PDB) and extracts content such as headers, metadata, and data samples to include in a generated markdown report. This creates a surface for indirect prompt injection where malicious content inside a scientific data file could attempt to influence the agent's behavior when it reviews the analysis results.
Ingestion points: The scripts/eda_analyzer.py script and the recommended custom analysis workflows ingest data from user-provided file paths using various scientific libraries like biopython, pandas, and h5py.
Boundary markers: The generated report uses markdown code blocks and sections to structure the output, but lacks explicit instructions to the agent to ignore potential commands embedded within the extracted data.
Capability inventory: The skill is capable of reading local files and writing reports to the filesystem, but does not have network access or arbitrary code execution capabilities.
Sanitization: Data extracted from the files is included in reports using standard formatting (like json.dumps for statistics) but without specific filtering for prompt injection patterns in text fields like headers, sequence IDs, or metadata fields.

exploratory-data-analysis