idea-discovery

To find and validate research ideas, this Skill orchestrates an end-to-end pipeline chaining literature review, idea generation, novelty verification, and critical review into ranked reports with pilot experiments.

You can generate research ideas from a reference paper by providing it during initialization. The workflow summarizes the paper first and steers idea generation to improve or extend that specific work.

The best way to plan pilot experiments during idea discovery is using a checkpointed validation workflow that enforces GPU and pilot budgets, running limited parallel pilots before refining the top proposal.

Yes, you can refine a research proposal into an experiment plan automatically. After deep-checking novelty, the workflow writes refined outputs for the top-ranked idea into a concrete next-step plan.

No, you do not need to run literature review and novelty verification separately for research planning. The pipeline chains these sub-skills together, executing the full discovery workflow from a single command.

Parallel pilot experiments are constrained by enforced pilot and GPU budgets to limit resource consumption. This checkpointed validation ensures only the top-ranked ideas undergo deep novelty checks and proposal refinement.