The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The scripts/new_notebook.py script takes a --title argument which is interpolated directly into a markdown cell in the generated Jupyter notebook. This creates a surface where a maliciously crafted title could include instructions designed to influence AI agents that read or execute the notebook later.
Ingestion points: The --title CLI parameter processed in scripts/new_notebook.py.
Boundary markers: None present. The input is placed directly into the notebook's markdown source list.
Capability inventory: The skill allows Write and Bash tools, enabling file creation and script execution. It also recommends installing jupyterlab for execution environments.
Sanitization: No sanitization or escaping is performed on the title string before it is written to the notebook JSON structure.
[COMMAND_EXECUTION]: The skill's primary workflow involves the agent executing a local Python script (scripts/new_notebook.py) using the uv run command. While intended for scaffolding, this involves passing user-influenced arguments to a shell command.
[EXTERNAL_DOWNLOADS]: The skill instructions suggest using the uv package manager to install jupyterlab and ipykernel from official Python registries to enable local execution of the generated notebooks.

openai-jupyter-notebook