run-experiment

Orchestrating experiment pipelines requires triggering Launch mode once all fidelity reports are ALL CLEAR. The skill then manages the full lifecycle, detaching long-running processes safely and writing structured result files for analyst interpretation.

The Phase 2 experiment lifecycle contract ensures runs are launched and finalized correctly using sentinel-based completion checks. It manages run and log operations while remaining backend-agnostic, detaching background execution safely.

Yes, experiment workflows support SSH, SLURM, local, and container backends. The skill abstracts execution details and remains backend-agnostic, allowing designated workflow runners to execute across multiple environments.

Generating structured result files happens automatically during the Finalize phase of experiment execution. The skill writes phase_2_verification/results/result_{n}.md files along with logs and sentinel checks upon run completion.

Yes, a designated runner is required to execute experiment workflows. The skill applies to environments where modules are ready, fidelity reports are ALL CLEAR, and a runner is available to handle the workflow.

You should not use this approach if fidelity reports are not ALL CLEAR or if modules are not ready. It requires a designated runner and only applies when the spec is ready and the environment supports safe background process detachment.