What problem does it solve? Research phases often start planning without clear methodological decisions, forcing downstream agents to guess or re-ask the user. This Skill captures physics and methodology decisions upfront in a CONTEXT.md file so researchers and planners can act without further clarification. ## Core Features & Use Cases - Gray Area Identification: Analyzes the phase from ROADMAP.md to surface phase-specific methodological ambiguities (formalism, approximations, boundary conditions, observables) rather than generic questions. - Socratic Dialogue: Conducts focused discussion loops that probe assumptions, question approximations, identify limiting cases, and surface anchors, with hard bounds on question rounds. - Structured CONTEXT.md Output: Writes decisions, physical assumptions, limiting behaviors, anchor registries, skeptical review items, and deferred ideas into a versioned phase file committed to git. - Use Case: Before planning a phase like "Compute Mott transition in Hubbard model", run this Skill to decide on impurity solver, cluster geometry, and transition identification criteria, producing a CONTEXT.md the planner consumes directly. ## Quick Start Ask the AI to run the gpd-discuss-phase skill for phase 2 to discuss methodological gray areas and generate the phase CONTEXT.md before planning.
