Adds a new data-science skill category with jupyter-live-kernel, which
uses hamelnb (https://github.com/hamelsmu/hamelnb) to give the agent
a live Jupyter kernel for stateful, iterative Python execution.
Key features:
- Variables persist across executions (unlike execute_code which is stateless)
- Inspect live variables, edit notebook cells, restart-and-run-all
- Clear trigger conditions and distinction from execute_code/terminal
- Practical tips based on hands-on testing
- No new tools required — uses terminal to run CLI commands
Prerequisites: uv, jupyterlab, hamelnb cloned to ~/.agent-skills/hamelnb
cfc3ccb212