What problem does it solve?
Going from a broad research direction to a validated, pilot-tested research idea requires many manual steps: surveying literature, brainstorming, checking novelty, and getting critical feedback. This Skill automates that entire pipeline so researchers can quickly identify ideas worth pursuing.
Core Features & Use Cases
- End-to-End Pipeline: Chains literature survey, idea generation, novelty verification, external review, and method refinement into one automated workflow producing a ranked IDEA_REPORT.md.
- Pilot Experiment Validation: Runs small GPU pilot experiments (with time and budget caps) on top ideas so ranking is based on empirical signal, not just theoretical appeal.
- Reference Paper Mode: Accepts an arXiv URL or local PDF as a base paper, summarizes it, and generates ideas that improve or extend it.
- Use Case: A PhD student types a direction like "efficient long-context attention" and receives a report with ranked ideas, confirmed novelty against recent arXiv work, reviewer scores, and a concrete experiment plan for the top idea.
Quick Start
Run the idea discovery pipeline on my research direction by saying: run the full idea discovery workflow for efficient long-context attention.
