What problem does it solve? Research sessions often stall when multiple plausible directions compete and no structured record explains why one was chosen. This Skill turns scout evidence into a traceable decision, comparing 2-3 candidate ideas on novelty, feasibility, evidence, cost, risk, and alignment so the next session can resume with full context. ## Core Features & Use Cases - Structured Candidate Comparison: Produces 2-3 testable directions, each with a mechanism hypothesis, closest alternative, minimum experiment, expected signal, cost, and risk. - Factor Card Validation: Creates and validates factor-card-<slug>.json decision artifacts via ds_lite_protocol.py validate-factor-card, keeping unmeasured factors unknown instead of inventing scores. - Graph-Tracked Decisions: Records branch, supersedes, and rollback edges in the research graph, writes an idea-<slug>.md artifact, and renders STATUS.md with the next experiment command. - Use Case: After a literature scout, you have three possible approaches to a problem. Use this Skill to score each on a Factor Card, pick the cheapest discriminating ablation, and log the decision so a future session knows exactly why that route was promoted. ## Quick Start Ask the assistant to propose 2-3 testable research directions from the latest scout artifact, compare their risks with Factor Cards, and select one route for immediate validation.
