What problem does it solve? Before planning a research phase, the AI often carries hidden assumptions about the physics, mathematics, numerics, and scope that may be wrong. This Skill makes those assumptions explicit so users can correct misconceptions early, before they propagate into plans and execution. ## Core Features & Use Cases - Nine-category assumption analysis: Surfaces physical, mathematical, approximation, computational, scope, anchor, expected-result, dependency, and user-binding assumptions, each with rationale and consequence-if-wrong. - Confidence and risk labeling: Marks assumptions as fairly confident, assumed, unclear, or risky, and names the weakest anchor, the earliest disconfirming check, and likely false progress. - Roadmap validation: Verifies the requested phase number against .gpd/ROADMAP.md and loads project state from .gpd/STATE.md before analyzing. - Use Case: A physics researcher about to plan Phase 3 of a turbulence study runs this Skill to see what regime, resolution, and benchmark anchors the AI assumes, then corrects a wrong coupling-regime assumption before any plan is written. ## Quick Start Ask the AI to list its assumptions for a specific phase, for example by invoking the skill with a phase number such as 3 after your roadmap exists in .gpd/ROADMAP.md.