What problem does it solve? Turning a vague request into working code often fails because the plan restates intent and pushes discovery onto whoever executes it. This Skill produces a plan a fresh session with zero context can execute: verified file paths, complete code per step, and a Run/Expected pair proving every change. ## Core Features & Use Cases - Dependency-ordered task planning: Draws order edges between tasks, splits shared write targets, and sizes each task to one commit with a Plan-task trailer. - Repository-grounded discovery: Builds a repository map of tracked files and exported symbols via repo-map.mjs, delegating deeper search to an explorer agent so the session context stays clean. - Plan validation: Runs plan-check.mjs against the finished plan to catch missing Commit blocks, git add mismatches, steps without Run/Expected, placeholders, and oversized tasks. - Use Case: Given an issue number or a shaped brief, produce a deliverable plan under docs/plans/ with full code for each step, so another session or a cheaper model can paste and run it without deciding anything. ## Quick Start Ask the AI to plan the work for a specific outcome, spec path, or issue number, for example: plan the implementation of issue 42 as a deliverable plan another session can execute.