What problem does it solve? Researchers with broad interests and constraints (e.g., CPU-only, theory-heavy, short timeline) often waste effort running full idea-generation pipelines on poorly scoped directions. This Skill pre-filters and crystallizes research directions into 1-3 concrete, well-scoped ideas before committing to a full pipeline. ## Core Features & Use Cases - Constraint-Driven Direction Discovery: Parses user constraints (resources, style, timeline, venue) and uses an external LLM plus web search to propose and validate 3-5 active research directions. - Deep-Dive Topic Generation & Novelty Check: Generates concrete paper topics per direction, runs quick arXiv novelty checks, and ranks topics across five scoring dimensions. - Idea Crystallization & Pipeline Prompts: Produces full idea cards (problem definition, theorem roadmap, minimal experiments, risk list, weekly plan) plus ready-to-paste prompts for a downstream idea-pipeline. - Use Case: A researcher says "low-resource theory work on Transformer mechanisms, no GPU, targeting NeurIPS" and receives two crystallized idea cards with executable pipeline prompts. ## Quick Start Ask the assistant to filter your research interests by stating your constraints and target venue, for example: filter ideas for CPU-friendly theory-heavy work targeting ICML.
