research

You can resume an interrupted experiment pipeline by using the `--start-from` flag. The workflow saves durable state in `pipeline-progress.md` after each stage, allowing you to safely recover and resume progress without losing completed work.

The best way to evaluate research claims against evidence is using a verdict-driven iteration loop. This Skill applies claim evaluation against experimental results and triggers at most one refined iteration loop if claims are insufficient, strictly capped by guardrails.

To generate a submission-ready paper from experimental results, you need an orchestration workflow that optionally produces the document. This Skill orchestrates the entire pipeline from idea discovery through experiment execution and verdict evaluation to optionally output a submission-ready paper.

Yes, you need YAML to define stages and gate logic for research orchestration. This Skill requires YAML-defined stages and gate logic to coordinate the end-to-end empirical research workflow and manage human gates across the pipeline.

To start research orchestration from a vague idea automatically, execute the pipeline with your research direction and the auto flag. This triggers end-to-end orchestration from empty wiki to paper-ready outputs, coordinating stages from idea discovery through experiment deployment.