workflows/python-analysis/Python analysis producing a traceable artifact — Run analysis code in the persistent Python kernel and register the result as a checksummed artifact. This is the most ordinary research run the system performs.
Steps: open_session, run_python_cell, save_artifact
Permissions: kernel:python, workspace:write
Declared artifacts: scores.csv
| File | Purpose |
|---|---|
workflow.json |
The versioned manifest: steps, permissions, declared artifacts, failure conditions, and the cases below. Version 1.0.0. JSON rather than YAML for the same reason the core is, and versioned because a benchmark whose cases can change silently measures nothing across time. |
| Case | Declared outcome | What it pins |
|---|---|---|
python-analysis/happy |
success |
Compute and write a CSV |
python-analysis/cell-error |
failure |
Analysis code that raises must fail, not produce silently |