The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides a structured pattern for natural language to code conversion in the nl_to_pandas function. This pattern explicitly includes steps for generating and executing pandas or SQL code based on user questions, which represents a dynamic execution surface that could be exploited if the natural language input is not properly sanitized.
[PROMPT_INJECTION]: The skill is designed to analyze external datasets from CSV, Excel, and database sources. It lacks implementation details or instructions for establishing security boundaries (such as delimiters or specific 'ignore' instructions) when processing this untrusted data, making the agent susceptible to indirect prompt injection attacks hidden within the data files.
[EXTERNAL_DOWNLOADS]: The documentation and code samples reference several external libraries and platforms including PandasAI, PyGWalker, Airbyte, and Airflow. While these are well-known data engineering tools, the skill encourages their integration and use for processing potentially untrusted datasets.

data-analysis