← 返回
未分类 中文

Writing Plans

Break work into clear, ordered plans before execution. Use when a task is multi-step, long-running, easy to derail, or needs coordination across tools, files...
在执行前将工作拆解为清晰有序的计划。适用于多步骤、耗时较长、易偏离方向或需要跨工具/文件协调的任务。
jisang1000 jisang1000 来源
未分类 clawhub v1.0.0 1 版本 100000 Key: 无需
★ 0
Stars
📥 513
下载
💾 15
安装
1
版本
#latest

概述

Writing Plans

Write plans that are concrete enough to execute, review, and verify.

Core Rule

Do not jump into a long task with only a vague idea.

Break it into explicit steps with outcomes.

When to Use

Use for:

  • multi-step setup work
  • migrations
  • skill installation and validation passes
  • browser automation workflows
  • refactors
  • research + implementation tasks
  • tasks that may require sub-agents

Plan Shape

A good plan includes:

  1. Goal
  2. Constraints
  3. Ordered steps
  4. Verification points
  5. Possible blockers

Good Step Format

Each step should answer:

  • what to do
  • what tool/file it touches
  • what counts as success

Example:

  1. Inspect installed files and runtime requirements.
  2. Install missing low-risk dependencies.
  3. Run smoke tests.
  4. Separate working / needs-auth / broken states.
  5. Report results.

Execution Rule

After writing the plan:

  • execute in order
  • update the plan if reality changes
  • do not hide changed assumptions

Lightweight vs Heavyweight

Lightweight plan

Use for 3-5 step tasks.

Heavyweight plan

Use when:

  • many files or tools are involved
  • external systems are involved
  • the task may run long
  • sub-agents may help

Reporting

Summarize with:

  • planned scope
  • what completed
  • what changed during execution
  • what remains

Practical Examples

Example: Skill setup pass

  1. Inspect files and dependencies
  2. Install missing low-risk runtime tools
  3. Run smoke tests
  4. Split results into working / needs-auth / blocked
  5. Report next steps

Example: Browser automation task

  1. Confirm target site and success condition
  2. Test a minimal path first
  3. Add login/session handling if needed
  4. Re-test the actual target flow
  5. Record blockers clearly

版本历史

共 1 个版本

  • v1.0.0 当前
    2026-03-30 19:48 安全 安全

安全检测

腾讯云安全 (Keen)

安全,无风险
查看报告

腾讯云安全 (Sanbu)

安全,无风险
查看报告

🔗 相关推荐

ai-agent

Agent Browser

rez0
用于 AI 代理的浏览器自动化 CLI。当用户需要与网站交互(包括浏览页面、填写表单、点击按钮、截图等)时使用。
★ 868 📥 348,366
ai-agent

Find Skills

root
帮助用户发现和安装智能体技能,当用户询问如「如何做X」、「找X的技能」、「有能做...的吗」等问题时
★ 1,521 📥 579,440
ai-agent

self-improving agent

pskoett
记录自身发现以实现自我改进的技能
★ 4,169 📥 941,724