The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to process external datasets, which is the primary and intended purpose of an EDA tool. While loading external files (e.g., data.csv) is a potential ingestion point for untrusted data, the skill demonstrates best practices by incorporating data validation frameworks.
Ingestion points: Data is loaded in SKILL.md and engineering/SKILL.md using pandas.read_csv and GreatExpectations.get_batch.
Boundary markers: Not explicitly defined in the provided code snippets, which is typical for documentation-focused examples.
Capability inventory: The skill uses visualization libraries (seaborn, matplotlib) and automated reporting tools (pandas_profiling) to generate insights and local HTML files.
Sanitization: The engineering guidelines explicitly recommend using Great Expectations for data validation (checking for nulls, uniqueness, and value ranges), which serves as a primary defense against malformed or malicious data inputs.
[EXTERNAL_DOWNLOADS]: The skill references established and well-known data science libraries and tools, including Pandas, Great Expectations, Datafold, and Amundsen. These are industry-standard resources used for data quality and discovery.

eda