What problem does it solve? Machine learning experiments often start without clear hypotheses or success criteria, making results hard to interpret. This Skill guides you through a structured interview to define purpose, verifiable hypotheses, and evaluation criteria before running an experiment, then saves the plan as a Markdown document. ## Core Features & Use Cases - Guided Interview: Asks one question at a time with recommended options and reasons, covering purpose, hypotheses, experiment design, implementation, and scope. - Standardized Plan Documents: Generates experiment plans from a template with metadata tables, hypothesis sections, and placeholders for post-experiment results, saved to .mjun/experiments/ with date-based filenames. - Verifiable Hypotheses: Ensures each hypothesis can later be judged as supported, partially supported, refuted, or unverifiable against agreed numeric criteria. - Use Case: Before training a new model variant, ask the Skill to draft an experiment plan; it interviews you about baselines and metrics, then produces a plan document you can link from Issues and PRs. ## Quick Start Ask the assistant to create a machine learning experiment plan for your next training or evaluation run.
