What problem does it solve? Research ideas often jump straight to experiment design without verifying whether the problem is worth solving, whether prior work already solved it, or whether the implementation chain is even feasible. This Skill grounds research directions in reality first, so experiments only run when they resolve a decision-relevant uncertainty. ## Core Features & Use Cases - Four-Stage Convergence Gate: Sequentially checks value (who is blocked, metric-to-goal chain), related-work reality (implementation-level facts via knowledge_keeper), end-to-end implementation feasibility (labeled arrow chain), and which single uncertainty to verify first. - Speculation Tracking: Maintains a speculation ledger scoring each unverified claim by guesswork level, failure impact, and verification cost to prioritize the cheapest decisive test. - Decision-Driven Experiments: Rejects experiments that do not target the top-ranked uncertainty, requiring pre-written actions for each possible outcome before anything runs. - Use Case: A researcher proposes applying causal intervention to radiology report generation. The Skill challenges the framing, delegates literature checks, maps the implementation chain, and outputs a verdict (rejected / blocked / evaluating / passed) with the single next decision-changing evidence item. ## Quick Start Ask the assistant to evaluate whether your proposed research direction is worth pursuing and what should be verified first.
