The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted scientific data files (e.g., FASTA, PDB, CSV) and extracts metadata such as headers, IDs, and attributes to include in generated markdown reports.
Ingestion points: Metadata extraction in scripts/eda_analyzer.py and recommended custom analysis steps in SKILL.md.
Boundary markers: None identified; extracted metadata is directly interpolated into markdown templates without specific delimiters or instructions to ignore embedded content.
Capability inventory: File system reading, report generation (writing to file system), and recommendations for package installation.
Sanitization: The script scripts/eda_analyzer.py lacks explicit sanitization or escaping of metadata extracted from files before placing it into the markdown report.
[EXTERNAL_DOWNLOADS]: The skill's documentation and reference files (e.g., references/bioinformatics_genomics_formats.md) recommend installing a large variety of domain-specific Python libraries via pip install or uv pip install.
The instructions explicitly suggest acquiring packages like biopython, rdkit, pysam, nd2reader, and many others based on the specific format being analyzed.
While most of these are standard tools in scientific computing, the lack of version pinning and the reliance on a vast number of external sources introduces a minor supply-chain risk factor.
[COMMAND_EXECUTION]: The SKILL.md instructions guide the agent to execute the scripts/eda_analyzer.py script and perform custom Python-based analysis involving file system operations.
The script performs file system reads and writes based on user-provided file paths.
There is a minor risk of command injection if the agent does not properly handle file paths containing shell metacharacters when invoking the script through a terminal environment.

exploratory-data-analysis