What problem does it solve? Research projects lose decisions, rationale, and execution order between sessions, forcing later work to reconstruct why choices were made or which plan is still live. This Skill produces a single PLAN.md that records decisions with rationale, pins dependency APIs, gates novelty claims, and stays unambiguous across pivots. ## Core Features & Use Cases - Structured plan authoring: Enforces a fixed section shape from framing through decision log, with pinned API surfaces, baselines, and a step-0 novelty literature-search gate. - Revision gate: Detects competing forward-looking plan sections by status marker rather than headings, ensuring exactly one live execution order survives any pivot. - Append-only decision log: Records hard-to-reverse decisions with rejected alternatives, parked/superseded statuses, and revisit conditions. - Use Case: After choosing a research direction, draft a PLAN.md that a future session can execute cold, then revise it in place when an experiment kills a branch, moving the dead branch's verdict to LOG.md. ## Quick Start Write a PLAN.md for my research project covering framing, decisions, pinned dependencies, experiments, execution order, and risks, starting with the step-0 novelty literature-search gate.
