The Agent Skills Directory

[DYNAMIC_EXECUTION]: The script scripts/sheetsmith.py uses pandas.DataFrame.eval() and pandas.DataFrame.query() with the engine="python" parameter in the run_transform and run_filter functions. This engine evaluates arbitrary Python expressions. While intended for data manipulation, this can be exploited to execute unauthorized code if an attacker can influence the expression string.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process external CSV and Excel files, which constitutes a vulnerability surface for indirect prompt injection.
Ingestion points: The load_dataframe function in scripts/sheetsmith.py reads data from files provided at runtime.
Boundary markers: None identified. The skill does not use delimiters or instructions to ignore embedded commands in the data.
Capability inventory: The skill can read/write files (load_dataframe, save_dataframe), drop/rename columns, and execute dynamic expressions via pandas.
Sanitization: No sanitization or validation of the input data or the resulting expressions is performed before execution.
[COMMAND_EXECUTION]: The test suite tests/test_sheetsmith.py utilizes subprocess.run to execute the CLI script during unit testing. While this is standard for testing, it is a mechanism for shell command execution.

sheetsmith