The Agent Skills Directory

[PROMPT_INJECTION]: The skill includes a prompt improvement hook (improve-prompt.py) that implements a bypass character (*). This allows users to explicitly skip the refinement process. This is an intended functional feature for user convenience and does not constitute a malicious attempt to subvert agent safety protocols.
[COMMAND_EXECUTION]: A Python script improve-prompt.py is provided as a pre-processing hook. The script processes user prompts by escaping special characters (backslashes and quotes) before embedding them into an evaluation template, which mitigates simple injection attacks during the refinement phase.
[DATA_EXFILTRATION]: The skill provides detailed workflows for processing various data formats (CSV, PDF, Excel, etc.). It explicitly mandates a 'Security Probe' step that limits data ingestion to metadata and small samples (e.g., 5 rows), which is a best practice for preventing data leakage and token overflow issues.
[EXTERNAL_DOWNLOADS]: The skill documentation references several standard, well-known data science libraries such as pandas, numpy, scikit-learn, matplotlib, seaborn, plotly, tensorflow, and pytorch. These are trusted industry-standard packages.

data-analysis