plan

Plan

Use this skill as the replacement for the deprecated custom prompt /prompts:plan.

Invocation

Preferred explicit trigger:

• $plan 帮我设计用户登录模块的实现方案

Natural-language triggers:

• 进入 Plan 模式
• 先出执行计划
• 帮我生成 plan 文件

Workflow

1. Confirm the task and current workspace.
2. Gather only the context needed to name concrete files, modules, or integration points.
   • Start with 5-8 targeted read-only tool calls.
   • Stop early once the implementation path is clear.
3. Choose simple, medium, or complex based on scope and risk.
4. If mcp__sequential-thinking__sequentialthinking is available and the task is more than trivial, use it to shape the plan.
5. Create or update a plan file under the current workspace.
   • New plan: plan/YYYY-MM-DD_HH-mm-ss-<slug>.md
   • Existing plan: reuse the previous file when the user clearly says they want to adjust the earlier plan.
6. The plan file frontmatter should include:

---
mode: plan
cwd: <current workspace>
task: <task summary>
complexity: <simple|medium|complex>
tool: mcp__sequential-thinking__sequentialthinking
total_thoughts: <final count>
created_at: <ISO8601 timestamp>
---

1. The plan body should contain:
   • A brief task summary
   • A numbered execution plan
   • Risks and blockers
   • path:line references where they materially help implementation
2. Reply in concise Chinese using this shape:
   • 🎯 任务
   • 📋 执行计划
   • 🧠 当前思考摘要
   • ⚠️ 风险与阻塞
   • 📎 Plan 文件

Defaults

• Prefer writing a plan/*.md file instead of only chatting.
• Keep the plan decision-complete so implementation can start without extra design decisions.
• Do not implement the plan unless the user explicitly switches from planning to execution.
• If file writes are blocked, return the full plan text plus the intended file path.