The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process .ipynb files, which are externally provided data sources. If these notebooks contain malicious instructions or hidden code, the agent may inadvertently execute them when performing operations like execution or conversion.
Ingestion points: The skill reads notebook content through nbformat.read(), jupytext, and Cursor's EditNotebook tool as described in SKILL.md.
Boundary markers: There are no explicit instructions or delimiters provided to the agent to treat notebook cell content as untrusted or to ignore embedded instructions.
Capability inventory: The skill provides multiple methods for code execution, including papermill shell and Python API execution, as well as jupyter nbconvert --execute commands.
Sanitization: No validation or sanitization of notebook cell content is mentioned before execution or processing.
[COMMAND_EXECUTION]: The skill provides numerous shell command templates for notebook manipulation.
Evidence: Instructions include python -m jupytext, papermill input.ipynb output.ipynb, and jupyter nbconvert --to html in SKILL.md.
[DYNAMIC_EXECUTION]: The skill facilitates the execution of notebook code at runtime through standard Jupyter ecosystem tools.
Evidence: The use of papermill.execute_notebook() and jupyter nbconvert --execute enables the programmatic execution of code contained within notebook cells.

jupyter-notebooks