What problem does it solve? Going from a broad research direction to a concrete, novel, and empirically validated research idea normally requires weeks of manual literature review, brainstorming, novelty checking, and pilot experiments. This Skill automates that entire pipeline, chaining literature survey, idea generation, novelty verification, external review, and experiment planning into one workflow. ## Core Features & Use Cases - End-to-End Pipeline: Chains literature survey, idea brainstorming, deep novelty checks, cross-model critical review, and method refinement into a single automated flow with checkpoints. - Pilot Experiment Validation: Runs up to 3 parallel GPU pilot experiments (with hour and budget caps) so ideas are ranked by empirical signal, not just theoretical appeal. - Structured Deliverables: Produces a canonical IDEA_REPORT.md with ranked ideas, plus a refined proposal, experiment plan, and research contract for downstream execution. - Use Case: A researcher types a direction like "efficient long-context attention" and receives a literature landscape, ranked pilot-tested ideas with novelty confirmation, reviewer scores, and a ready-to-run experiment plan. ## Quick Start Run the idea discovery pipeline on your research direction, for example by asking to start the full idea discovery workflow for efficient long-context attention mechanisms.
