What problem does it solve? Complex tasks that span multiple context windows or sessions lose all progress when a conversation ends, because only the filesystem survives between sessions. This Skill turns large tasks into version-controlled execution plan files so work can be interrupted, handed off between agents, and resumed without losing goals, decisions, or acceptance criteria. ## Core Features & Use Cases - Versioned Plan Artifacts: Creates structured plan files under docs/exec-plans/active/ with goal, scope, non-goals, steps, decision log, acceptance criteria, and risks, then moves them to completed/ instead of deleting them. - Plan Quality Enforcement: Rejects plans whose acceptance criteria are not mechanically verifiable, whose steps cannot be self-verified within one PR, or that are edited concurrently by multiple agents. - Multi-Agent Handoff: Uses the active/ directory as a shared coordination ledger with owner markers, decision logs, and a tech-debt tracker for blocked or deferred work. - Use Case: A user asks to migrate a frontend build from Webpack to Turbopack across several sessions with multiple agents. The Skill creates docs/exec-plans/active/frontend-build-migration.md, decomposes it into independently verifiable steps, records decisions, and hands off to a verification loop for step-by-step execution. ## Quick Start Ask the agent to create an execution plan for a large refactoring task that will span multiple sessions, and it will write a structured plan file under docs/exec-plans/active/ with verifiable steps and acceptance criteria.
