explore-run

What problem does it solve?

This skill helps researchers plan and summarize isolated exploratory runs for deep learning projects, ensuring explicit authorization and safe separation from trusted baselines.

Core Features & Use Cases

• Plan and rank exploratory variants for small-subset validations, quick-cycle sweeps, or idle-GPU searches, with results stored in explore_outputs/ (CHANGESET.md, TOP_RUNS.md, status.json).
• Integrate with execution policy guidance in references/execution-policy.md and use plan_variants.py and write_outputs.py to generate and publish results.
• Use-case: when you need candidate-only exploration results that can be reviewed before any trusted deployment.

Quick Start

Plan an explicit exploratory run for the current research task and generate the explore_outputs bundle with CHANGESET.md, TOP_RUNS.md, and status.json.