name: experiment-provenance description: Capture experiment provenance with reproducible run metadata, artifact pointers, and decision logs for scientific claims.
Experiment Provenance
Use this skill when a task produces evidence that may become a manuscript claim.
When to use
- "track provenance"
- "log this experiment"
- "record reproducible run details"
- "capture evidence for manuscript claim"
Required provenance fields
- Objective and hypothesis.
- Exact command(s) executed.
- Input files, config, and environment assumptions.
- Output artifact paths.
- Validation status (tests/build/render).
- Claim boundary and uncertainty notes.
Workflow
- Record run intent before execution.
- Execute with deterministic commands where possible.
- Store output artifact paths, not just summaries.
- Link provenance notes to
plans/main-plan-activity.mdandplans/main-plan-log.csv. - Mark evidence quality: strong, partial, or insufficient.