The Agent Skills Directory

[PROMPT_INJECTION]: The skill uses multiple user-controlled arguments ($PRODUCT, $DATASET_TYPE, $COLUMNS, $CONSTRAINTS) which are interpolated into the final output. This presents a surface for indirect prompt injection where malicious values in the data fields could attempt to influence the logic of the generated content.
[COMMAND_EXECUTION]: The skill is designed to generate executable Python scripts as a primary deliverable. Evidence: The template in SKILL.md uses string interpolation to set script variables (e.g., FILENAME = "$DATASET_TYPE.csv"). Without sanitization of the $DATASET_TYPE argument, an attacker could provide input that breaks the string literal to inject arbitrary Python code into the generated file.

dummy-dataset