AgentSkillsCN

ideation-bootstrap-planner

将产品或应用创意转化为适用于该堆栈模板单体仓库的专属实施计划(Bun工作空间搭配Vite React前端、Hono后端、共享包以及Firebase)。当用户表示想要构建某项产品或应用时(例如,“构建XYZ”),或需要启动引导指导,或希望获得分阶段执行规划时,可使用此功能。仅在必要时提出澄清性问题;当领域或集成尚不明确时,应展开针对性研究;同时,生成PLAN.md文件,并依据此仓库结构映射任务列表。

SKILL.md
--- frontmatter
name: ideation-bootstrap-planner
description: Convert a product/app idea into a repo-specific implementation plan for this stack-template monorepo (Bun workspace with Vite React frontend, Hono backend, shared package, Firebase). Use when a user says they want to build something (for example, "build xyz"), needs bootstrapping guidance, or asks for phased execution planning. Ask clarifying questions only when required, perform focused research when the domain or integrations are unclear, and generate PLAN.md plus tasks.json mapped to this repository structure.

Turn a raw idea into a practical execution plan for this repository.

Core Outcome

Produce two files at project root:

  • docs/ideation/PLAN.md: phase-by-phase implementation plan.
  • docs/ideation/tasks.json: machine-readable task list grouped by phase.

Use this repository's architecture and commands, not generic templates.

Required Workflow

  1. Confirm scope from user input
  • Extract goal, target users, core workflow, and non-goals.
  • Record assumptions if input is incomplete.
  1. Ask only required clarifying questions
  • Ask questions only when missing answers materially change architecture or phase order.
  • Ask at most 5 concise questions in one round.
  • Prefer closed choices when possible.
  • If user does not answer, continue with explicit assumptions.
  1. Load repository context
  • Read AGENTS.md, README.md, root package.json, frontend/package.json, and server/package.json.
  • Align planning to:
    • frontend/ for UI, pages, state, API clients, and tests.
    • server/ for Hono routes, middleware, services, and tests.
    • shared/ for shared types and config.
  1. Perform focused research when needed
  • Do local repo research first.
  • Do web research if the idea depends on unfamiliar libraries, third-party APIs, security/compliance requirements, or complex domain rules.
    • Use context7 MCP server to read docs of frameworks if needed
  • Capture findings as short bullets with links inside PLAN.md under Research Notes.
  • Distinguish confirmed facts from assumptions.
  1. Design phased implementation
  • Break work into ordered phases where each phase can be validated.
  • Keep phases outcome-driven.
  • Define dependencies and risk per phase.
  • Ensure each phase maps to concrete folders/files in this repo.
  1. Generate PLAN.md
  • Use references/plan-template.md.
  • Keep language specific and implementation-focused.
  • Include architecture decisions for frontend, server, shared, and Firebase usage.
  • Include testing, observability, and rollout strategy.
  • Include testing tasks for both happy path and key error paths.
  1. Generate tasks.json
  • This is an array of tasks that a user or AI needs to perform step by step in order to complete a plan phase
  • Use template from tasks-template.json
  • Convert each phase into actionable task objects.
  • Include acceptance criteria and dependencies per task.
  1. Run quality gate before finalizing
  • Verify PLAN.md and tasks.json are consistent and phase names match exactly.
  • Ensure no phase is missing test tasks.
  • Ensure commands use Bun and existing workspace scripts.
  • Ensure unknowns are captured in Open Questions rather than hidden.

Planning Rules

  • Prefer simple vertical slices before broad platform work.
  • Plan MVP-first; defer optional enhancements to later phases.
  • Keep each phase independently reviewable.
  • Do not include code unless user explicitly asks.
  • Do not invent APIs, environment variables, or infra details without labeling assumptions.

Output Contract

PLAN.md must include:

  • Problem Statement
  • Solution Overview
  • Assumptions and Constraints
  • Repo Impact Map
  • Phase Plan
  • Testing Strategy
  • Risks and Mitigations
  • Research Notes
  • Open Questions

tasks.json must include:

  • Top-level metadata: idea, generatedAt, repo, phases
  • Phase entries matching PLAN.md names
  • Task objects with required fields:
    • id (number)
    • status (pending, done)
    • title (task title)
    • description (detailed task context)
    • notes (any additional notes from previous tasks)

Task object example:

json
{
  "id": "task-001",
  "status": "todo",
  "title": "Create todo route",
  "description": "Add Hono endpoints for todo CRUD with validation and error handling.",
  "notes": ["Endpoints should be under /api/"]
}

Load these references when needed:

  • plan-template.md
  • tasks-template.json