AgentSkillsCN

plan-with-agent

具备持久化页面状态的浏览器自动化。当用户希望浏览网站、填写表单、截取屏幕截图、提取网页数据、测试 Web 应用，或自动化浏览器工作流时，可优先选用此技能。触发短语包括：“前往 [URL]”、“点击”、“填写表单”、“截屏”、“抓取数据”、“自动化操作”、“测试网站”、“登录”等，或任何涉及浏览器交互的请求。切勿用于：分享 URL、嵌入链接、生成屏幕截图文件。

SKILL.md
--- frontmatter
name: plan-with-agent
description: "Creates implementation plans from ideas and requirements, generating Plans.md ready for /work. Use when user mentions '/plan-with-agent', create a plan, organize tasks, feature planning, or TDD planning. Do NOT load for: implementation, reviews, or setup."
allowed-tools: ["Read", "Write", "Edit", "Grep", "Glob", "Bash", "WebSearch", "Task"]
argument-hint: "[--ci]"

Planning Skill

Organizes ideas and requirements, converting them into executable tasks in Plans.md.

Quick Reference

  • •"Create a plan" → this skill
  • •"Turn what we talked about into a plan" → extract from conversation
  • •"Want to organize what to build" → start with hearing
  • •"Plan with TDD" → force TDD adoption
  • •"CI/benchmark" → --ci mode

Deliverables

  • •Plans.md - Task list executable with /work (required)
  • •Feature priority matrix - Required/Recommended/Optional classification

Usage

bash
/planning              # Interactive planning
/planning --ci         # CI mode (non-interactive)

Mode-specific Usage

ModeCommandDescription
Solo mode/planningClaude Code alone: plan → execute → review
2-agent mode/plan-with-cc (Cursor)Plan with Cursor → Execute with Claude Code

Feature Details

FeatureReference
Execution FlowSee references/execution-flow.md
TDD AdoptionSee references/tdd-adoption.md
Priority MatrixSee references/priority-matrix.md

Execution Flow Overview

  1. •Step 0: Check conversation context (extract from previous conversation or start fresh)
  2. •Step 1: Hearing what to build
  3. •Step 2: Increase resolution (max 3 questions)
  4. •Step 3: Technical research (WebSearch)
  5. •Step 4: Extract feature list
  6. •Step 5: Create priority matrix (Required/Recommended/Optional)
  7. •Step 5.5: TDD adoption judgment and test design
  8. •Step 6: Effort estimation (reference)
  9. •Step 7: Generate Plans.md with quality markers
  10. •Step 8: Guide next actions

Auto-invoke Skills

SkillWhen to Call
setupCall first (executes adaptive setup)
vibecoder-guideWhen user is non-technical

Quality Markers (Auto-assigned)

Task ContentMarkerEffect
Auth/login feature[feature:security]Security checklist
UI component[feature:a11y]a11y check
Business logic[feature:tdd]TDD recommended
API endpoint[feature:security]Input validation
Bug fix[bugfix:reproduce-first]Reproduction test first

Next Actions

After planning:

  • •Start implementation with /work
  • •Or say "start from Phase 1"
  • •Adjust with "add {{feature}}" or "postpone {{feature}}"