What problem does it solve?
Plan-writer stops teams from manually composing solution proposals by orchestrating a full workflow that respects requirements, knowledge, and output constraints so the AI never exposes the full plan to the user until the final file is written. It enforces mode detection, knowledge-base sourcing, draft critique, and humanizer polishing so every version stays aligned with customer needs and internal policies.
Core Features & Use Cases
- Mode-aware orchestration: Detects cold starts, iterative feedback, or dialog-only inputs, then triggers the right phases including pre-checks, draft creation, and critique loops.
- Knowledge and critique loops: Calls KB match once, drafts per template, and refines the content through a specialist agent before committing the plan.
- Controlled output and humanizer: Writes only to plan files, blocks plan snippets in chat, runs humanizer-zh on narrative sections, and reports status without leaking the solution.
- Use Case: When a solution team receives updated requirements, asking Claude to run plan-writer yields a polished solution version with version tracking, constraints awareness, and a final write into docs/plan/solution.md.
Quick Start
Ask Claude to run plan-writer with the latest requirements so it can generate and finalize the customer solution plan through the standard phases.