agentic-in Jupyter Live Kernel
Use this built-in skill when the user wants interactive notebook analysis, exploratory data work, or a live kernel workflow.
Core rules
- Confirm the runtime, dataset location, and dependency posture before executing notebook cells.
- Keep exploratory work reproducible enough to replay outside the current kernel.
- Separate quick investigation from durable scripts, tests, or pipelines.
- Record assumptions about data freshness, sampling, and environment state.
Default workflow
- Inspect the data source, schema shape, and available runtime.
- Load the smallest slice that can answer the question.
- Iterate interactively while keeping cells and outputs interpretable.
- Promote stable logic into scripts or documented procedures when the work becomes durable.
Guardrails
- Do not hide environment-specific state inside unexplained notebook magic.
- Do not treat one kernel run as a reproducible result by default.
- Do not keep long-lived production logic trapped in ad hoc cells.