The Agent Skills Directory

[SAFE]: The skill is a legitimate tool for scientific data analysis. The code provided in scripts/eda_analyzer.py is well-structured, follows security best practices (such as using the pathlib module for file path handling and sampling large files), and stays strictly within its defined scope.
[COMMAND_EXECUTION]: The skill executes a local Python script (scripts/eda_analyzer.py) to perform analysis. This script uses established scientific libraries to process data. The execution is localized to the user's environment and is triggered by specific user requests to analyze data files.
[PROMPT_INJECTION]: As the skill is designed to ingest and process untrusted external data files (scientific data from various sources), it possesses an inherent surface for indirect prompt injection. Maliciously crafted data files containing natural language instructions could attempt to influence the agent's behavior during the report generation phase. However, the skill implements basic sanitization by using structured JSON output for metadata and sampling methods to limit exposure.
Ingestion points: Scientific data files provided by the user via file paths (processed in scripts/eda_analyzer.py).
Boundary markers: The skill uses markdown headers and code blocks in its report generation, which helps separate data from instructions, though it does not explicitly warn the LLM to ignore instructions within the data.
Capability inventory: The skill can read local files, write markdown reports to the file system, and perform complex data processing using specialized libraries.
Sanitization: The analyzer script uses json.dumps for embedding metadata into reports, providing a layer of data structuralization.

exploratory-data-analysis