What problem does it solve?
Turning a high-level requirement into an actionable, dependency-ordered execution plan is slow and error-prone. This Skill decomposes requirements into a human-readable roadmap.md plus a machine-executable issues.jsonl, so planning output flows directly into execution pipelines without manual reformatting.
Core Features & Use Cases
- Dual Decomposition Modes: Progressive mode (MVP → Usable → Refined → Optimized layers) for high-uncertainty work, or Direct mode (topological task sequence) for well-defined requirements, with automatic strategy recommendation based on uncertainty assessment.
- Iterative Refinement with Decision Recording: Multi-round user feedback loops (up to 5 rounds) with cumulative context, optional web research for architecture patterns, and a structured decision trail recorded in the roadmap.
- Issue Store Integration: Creates issues in a global .workflow/issues/issues.jsonl file with wave tags, dependencies, priorities, and lifecycle requirements, queryable via ccw issue commands and consumable by csv-wave-pipeline.
- Use Case: Given a requirement like "Implement user authentication with OAuth and 2FA", the Skill assesses uncertainty, optionally explores the codebase via a subagent, decomposes the work into waves of issues with testable convergence criteria, refines the plan with the user, and hands off to an execution pipeline.
Quick Start
Ask the AI to plan a roadmap for your requirement, for example: run roadmap-with-file with the requirement "Implement user authentication system with OAuth and 2FA" and confirm the recommended decomposition strategy.