The Agent Skills Directory

[SAFE]: The skill provides instructional methodology and templates for data science tasks. No malicious scripts, hardcoded credentials, or unauthorized network operations were detected.
[PROMPT_INJECTION]: The skill's design involves ingesting and analyzing external datasets, which presents an attack surface for indirect prompt injection. 1. Ingestion points: User-provided datasets are processed during Stage 1 (Quality) and Stage 2 (EDA) as described in the guides. 2. Boundary markers: The templates use standard Markdown structure but do not explicitly instruct the agent to use delimiters or sanitization for untrusted data values. 3. Capability inventory: The workflow encourages the generation and execution of Python code for analysis and visualization (Stage 5). 4. Sanitization: The provided guides focus on data quality metrics rather than security validation or escaping of input data.

data-analysis