The Agent Skills Directory

[SAFE]: The skill is well-architected for its primary purpose of scientific data exploration. It provides clear instructions for file type detection and utilizes a standalone Python script to handle complex data parsing locally.
[COMMAND_EXECUTION]: The provided Python script eda_analyzer.py implements safe file operations using standard libraries like pathlib and os. It focuses on extracting metadata (size, modified date) and statistical aggregates (mean, std, shape) rather than executing arbitrary commands or user-supplied shell scripts.
[EXTERNAL_DOWNLOADS]: The skill documentation and reference files mention numerous well-known and reputable scientific libraries (e.g., NumPy, Pandas, Biopython, RDKit, MDAnalysis). These are standard dependencies for the scientific domain, and the skill correctly instructs the user/agent on their installation through standard package managers.
[DATA_EXFILTRATION]: No network operations, data exfiltration patterns, or hardcoded credentials were detected. The skill performs all data processing on local file paths provided by the user.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted scientific data files (e.g., CSV, JSON, FASTA). While this constitutes an attack surface, the risk is mitigated because the analysis script focuses on structural metadata and statistical summaries. It does not treat the content of the data files as natural language instructions, thereby reducing the likelihood of indirect prompt injection affecting the agent's behavior.

exploratory-data-analysis