The Agent Skills Directory

[SAFE]: The skill documents standard operational procedures for developers working on the pymc-marketing library. It includes instructions for running a local mock runner, executing notebooks for functional validation, and using linting tools.
[COMMAND_EXECUTION]: The skill uses uv run to execute various local scripts and tools (jupyter, ruff, pre-commit, gh). These commands are scoped to the development environment and are necessary for the skill's stated purpose of fixing and validating notebooks.
[INDIRECT_PROMPT_INJECTION]: The skill processes Jupyter notebook files (.ipynb) and executes their content via jupyter execute. While this represents an ingestion surface for untrusted data if a notebook were maliciously modified, the context is focused on internal repository maintenance where the agent is fixing failing notebooks already present in the codebase. The risk is minimized by the intended development workflow.
[DYNAMIC_EXECUTION]: The skill involves the execution of notebook cells and short Python validation scripts. This dynamic execution is the primary function of the skill (validating runtime behavior) and is performed within the controlled environment of the user's project.

working-with-notebooks