doc-writing_skill

Overview

This skill helps you write complete, standardized, and actionable requirement and technical documents, including PRDs, technical designs, API specs, database schemas, and AI feature designs. It supports both progressive (discuss-and-iterate) and one-shot (full draft) modes to match how well requirements are defined. The output follows clear templates, version tracking, and change markers to keep collaboration traceable.

How this skill works

The skill creates document skeletons from templates and guides the conversation to fill missing 5W2H items (What, Why, Who, When, Where, How, How Much). In progressive mode it incrementally updates the document with markers for additions, updates, deletions, and bug fixes and maintains a version history. It can generate PRD, technical solution, API design, database schema, AI feature design, competitor analysis, flow diagrams (Mermaid), and task breakdowns based on input.

When to use it

• You need a formal PRD or technical solution before development starts
• Requirements are unclear and you want to iterate while discussing
• You must produce API or database design for engineering handoff
• You are designing an AI feature with model, prompt, and cost analysis
• You need a traceable versioned document for review and approvals

Best practices

• Start with 5W2H: require explicit answers for What/Why/Who before drafting
• Choose document type based on scope: PRD for product, technical solution for architecture, API doc for implementation
• Use MoSCoW to lock MVP scope and avoid scope creep
• Keep incremental markers ([NEW], [UPDATED], [BUGFIX], deletion strike) during discussion and remove them at confirmation
• Include success metrics, non-functional requirements, dependencies and rollback/monitoring plans in every final doc

Example use cases

• Create a PRD for a new mobile onboarding flow with measurable success metrics
• Draft a technical solution for a payment microservice: architecture, DB design, API list, and rollout plan
• Design REST APIs: endpoints, request/response schemas, error codes and examples for frontend/backend teams
• Iteratively define an AI assistant feature: model selection, prompt templates, performance targets and cost estimates
• Produce a database design and migration plan with ER diagram, indexes and rollback scripts