The Agent Skills Directory

[COMMAND_EXECUTION]: The skill documents methods for programmatic execution of code and notebooks using nbconvert, nbclient, and papermill. This functionality allows an agent to execute arbitrary Python or shell code within the local environment via Jupyter kernels.
[EXTERNAL_DOWNLOADS]: Includes instructions for installing common data science libraries from public registries, such as jupyter, papermill, and pandas.
[DATA_EXFILTRATION]: Contains instructions for starting a Jupyter server bound to all network interfaces (--ip 0.0.0.0). While standard for remote access, this configuration permits remote connections and can lead to unauthorized data access or environment compromise if the server's authentication mechanisms (tokens/passwords) are not properly configured.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it instructs the agent to read and process external data files (e.g., data.csv) and notebook files (e.g., input.ipynb) without implementing boundary markers or sanitization to prevent embedded instructions from influencing the agent's behavior.
Ingestion points: pd.read_csv('data.csv'), nbformat.read(f), and papermill input.ipynb in SKILL.md.
Boundary markers: Absent. The instructions do not specify using delimiters or warnings to ignore instructions embedded in the processed data.
Capability inventory: The skill has the capability to write files (nbformat.write), execute shell commands (!ls, %%bash), and run arbitrary Python code via Jupyter kernels.
Sanitization: Absent. There is no mention of validating or filtering the content of the notebooks or data files before execution.

jupyter-live-kernel