Token导航 LogoToken导航

Agent Skill · 研究检索

plan-first(首先计划)

plan-first 用于查找、检索和筛选相关信息,适合在 OpenClaw 中需要根据关键词、任务场景或来源线索快速定位候选结果时使用。可结合来源仓库、安装命令和原始 README 继续核验具体用法。安装前建议确认权限范围、维护状态,以及是否会触发联网、命令执行或文件读写。

plan-first 是研究检索类 Agent Skill,可用宿主为 OpenClaw,通过ClawHub获取,权限为需要联网;许可证为 MIT-0,第三方扫描已通过,按安装量在研究检索类的 73,675 个 Skill 里排第 29,016。

总安装 3,127周安装 129GitHub Stars 公开资料未说明下载 1,022核验 2026-05-01

安装说明

本站只整理中文说明,不托管安装包,也不代用户安装。
GitHub
许可证
MIT-0
版本
1.0.0

安装方式

通过对话安装

复制提示词发给支持本地命令或 Skills 的 AI 助手,先确认命令和权限,再让它执行。

请帮我安装这个 Agent Skill:plan-first(首先计划)
来源仓库:https://github.com/simoncatbot/plan-first
安装命令:
openclaw skills install plan-first
安装前请先检查当前环境是否支持对应 CLI,并向我确认将要执行的命令、安装目录、联网范围和文件读写权限;确认后再执行。

命令行安装

复制命令到本机终端执行。该命令会通过 OpenClaw 从第三方来源获取 Skill;本站只展示命令,不托管安装包,也不自动执行。

ClawHubOpenClaw
openclaw skills install plan-first

这个 Skill 是做什么的

研究检索类 Skill 主要解决「找资料、核对事实、把来源整理清楚」的问题,通常会让助手先检索再总结。

Plan-First在行动前强制生成详细执行计划再推进任务。适合多步骤复杂项目分解为可验证子目标的管理模式。通过预设检查点控制实施节奏并记录过程日志。适用宿主包括 OpenClaw,接入前应确认版本、权限和运行环境要求。计划生成依赖用户输入完整性,模糊需求可能导致偏差。

plan-first 属于研究检索类 Skill,可作为该场景下的辅助能力补充。

上手步骤

4 步
  1. 1

    确认自己的宿主工具

    公开资料标注支持 OpenClaw;先确认自己用的是其中之一,再按对应工具的 Skills 目录放置。

  2. 2

    拿到安装命令

    可以直接复制的命令是 openclaw skills install plan-first;命令里的占位符换成自己的路径或密钥再执行。

  3. 3

    安装后重启宿主

    命令执行完要完全退出宿主工具再打开,让新装的 Skill 被重新扫描到,否则列表里可能看不到。

  4. 4

    按最小权限试用

    它需要需要联网,先在测试目录或测试账号里跑,确认没问题再放到正式项目。

常见问题

6 条

plan-first 怎么安装?

在终端执行 openclaw skills install plan-first。装完完全退出宿主工具再打开,让它重新扫描本地 Skill。

plan-first 支持哪些 AI 工具?

公开资料标注支持 OpenClaw。其他工具只要能读取本地 Skill 目录,一般也能手动放进去,但要自己确认触发方式是否一致。

plan-first 需要什么权限?

权限级别是需要联网。该 Skill 可能需要联网访问来源站点、仓库或外部 API;具体网络访问范围需要结合源码和 README 复核。

plan-first 是免费的吗?

来源标注的许可证是 MIT-0。本站不代收任何费用,也不托管安装包;能否商用、要不要署名,以许可证原文为准。

plan-first 用的人多吗?

公开数据是累计安装 3,127 次,近一周安装 129 次,下载 1,022 次。这些数字来自第三方目录统计,只能当作热度参考,不代表质量或安全性。

plan-first 安全吗?

第三方扫描结果为通过,但这只覆盖已知的静态检查项,不能替代自己读一遍源码和安装命令。

值得选的地方和要先想清楚的地方

值得选的地方

  • 专注 OpenClaw 一种宿主,触发方式和使用习惯不用来回切换。
  • 许可证明确为 MIT-0,商用和分发边界好判断。
  • 第三方扫描已通过,常见的高风险行为没有命中。

先想清楚的地方

  • 第三方目录整理,版本更新和行为细节以来源仓库为准。

适不适合你

更适合

  • 已经在用 OpenClaw,想直接扩展它能力的人。
  • 需要处理研究检索类任务、又不想每次都手写提示词的人。
  • 需要许可证清楚、并且有第三方扫描结果的团队项目。

不太适合

  • 需要厂商级技术支持和 SLA 的生产系统。

研究检索类 Skill 对比

6 个同类
Skill总安装GitHub 星权限许可证扫描
find-skills

寻找技能 · Codex、Claude

31,199,81316,486操作浏览器MIT有提醒
self-improving-agent

自我改进 Agent · OpenClaw

16,464,0023,396接触敏感数据MIT-0通过
soultrace

灵魂追踪 · Codex、Claude

7,588,94412需要联网未明确有提醒
self-improving

自我完善 · OpenClaw

6,829,2841,046需要联网MIT-0通过
microsoft-foundry

微软代工厂 · Codex、Claude

6,614,144764调用外部服务未明确有提醒
plan-first

首先计划 · 本页

3,127—需要联网MIT-0通过

按安装量取同分类前几名;数字来自第三方目录统计,只作热度参考,不代表质量结论。

SKILL.md 原文

name
plan-first
description
Solve complex multi-step tasks by generating a detailed plan before execution. Based on the Plan-and-Solve prompting research, this skill breaks tasks into clear steps, validates the approach, then executes systematically. Use for any task with multiple steps, dependencies, or where jumping straight to the answer might miss important details. Critical for coding, writing, analysis, and problem-solving tasks.

Plan First - Structured Problem Solving

Plan First implements the Plan-and-Solve research paradigm for tackling complex tasks systematically.

The core insight: Generative AI performs better when it plans before acting, especially on multi-step tasks where dependencies exist between steps.

The Analogy: Architect vs. Builder

Builder ApproachArchitect Approach (Plan First)
Start building immediatelyDesign blueprints first
Discover problems mid-projectAnticipate issues before they arise
May need to tear down and rebuildBuild correctly the first time
Works for simple structuresEssential for complex projects

When to Use Plan First

Perfect for:

  • 🏗️ Multi-step coding tasks (implement feature X that requires Y, Z, W)
  • 📝 Complex writing (reports, documentation with sections)
  • 🔬 Analysis tasks (data analysis with multiple phases)
  • 🧩 Problem-solving (math, logic, puzzles with steps)
  • 📋 Project planning (breaking down large initiatives)

Skip for:

  • ⚡ Single-step tasks ("What is 2+2?")
  • 🎯 Direct questions ("What is the capital of France?")
  • 🔄 Pure exploration (brainstorming without structure)

The Workflow

Step 1: Generate the Plan

Before ANY execution, create a detailed plan:

TASK: "Build a user authentication system"

PLAN:
1. [SETUP] Initialize database schema for users table
   - Dependencies: None
   - Success criteria: Migration runs without errors

2. [IMPLEMENTATION] Create password hashing utilities
   - Dependencies: Step 1 complete
   - Success criteria: bcrypt integration works

3. [IMPLEMENTATION] Build login endpoint
   - Dependencies: Steps 1, 2
   - Success criteria: Returns JWT on valid credentials

4. [IMPLEMENTATION] Build signup endpoint
   - Dependencies: Steps 1, 2
   - Success criteria: Creates user, returns token

5. [INTEGRATION] Add middleware for protected routes
   - Dependencies: Steps 3, 4
   - Success criteria: Unauthorized requests rejected

6. [TESTING] Write comprehensive tests
   - Dependencies: Steps 3, 4, 5
   - Success criteria: All tests pass

Step 2: Validate the Plan

Self-check before proceeding:

  • [ ] Are steps in logical order?
  • [ ] Are dependencies clearly marked?
  • [ ] Is each step specific and actionable?
  • [ ] Are success criteria verifiable?
  • [ ] Did I miss any prerequisite steps?

If gaps found: Revise the plan first.

Step 3: Execute Step by Step

Execute ONE step at a time, in order:

Executing Step 1: Initialize database schema
✓ Run: CREATE TABLE users...
✓ Verify: Table exists with correct columns
✓ Mark complete

Executing Step 2: Create password hashing utilities
✓ Run: Implement hash_password(), verify_password()
✓ Verify: Unit tests pass
✓ Mark complete

... continue until all steps complete

Step 4: Handle Deviations

If a step fails:

  1. Analyze why it failed
  2. Revise plan if needed (add missing steps, reorder)
  3. Resume from current step

If requirements change:

  1. Stop execution
  2. Revise remaining plan
  3. Continue with updated plan

Plan Format Template

## Task
[Clear statement of what needs to be done]

## Plan

### Step N: [Action Verb] [What to do]
**Dependencies:** [List prerequisite steps]
**Approach:** [How to accomplish this step]
**Verification:** [How to confirm success]
**Files/Tools:** [What to modify or use]

---

### Step 1: [SETUP/IMPLEMENTATION/ANALYSIS/TESTING] Description
**Dependencies:** None
**Approach:** 
- Detail 1
- Detail 2
**Verification:** [Specific test or check]
**Files:** [Files to create/modify]

### Step 2: ...
[Continue for all steps]

## Risk Analysis
**Potential blockers:** [What could go wrong]
**Mitigation:** [How to handle issues]

## Rollback Plan
**If fails at step N:** [How to undo/reset]

Examples

Example 1: Code Implementation

Task: "Add email verification to signup flow"

## Plan

### Step 1: [SETUP] Create email_tokens table
**Dependencies:** None
**Approach:** 
- Create migration for email_tokens table
- Fields: id, user_id, token, expires_at, used
**Verification:** Migration runs, table structure correct
**Files:** migrations/003_email_tokens.sql

### Step 2: [IMPLEMENTATION] Add email sending utility
**Dependencies:** Step 1
**Approach:**
- Create utils/email.py with send_verification_email()
- Use SMTP or SendGrid
- Include token in email link
**Verification:** Can send test email successfully
**Files:** utils/email.py

### Step 3: [IMPLEMENTATION] Modify signup endpoint
**Dependencies:** Steps 1, 2
**Approach:**
- After user creation, generate token
- Send verification email
- Return "check your email" response
**Verification:** Signup creates token and sends email
**Files:** routes/auth.py

### Step 4: [IMPLEMENTATION] Create verify endpoint
**Dependencies:** Step 1
**Approach:**
- GET /verify-email?token=xxx
- Check token valid and not expired
- Mark user as verified
**Verification:** Clicking link marks user verified
**Files:** routes/auth.py

### Step 5: [INTEGRATION] Add verified check to login
**Dependencies:** Step 4
**Approach:**
- Modify login to require verified=True
- Return appropriate error if unverified
**Verification:** Unverified users can't login
**Files:** routes/auth.py

### Step 6: [TESTING] Write tests
**Dependencies:** Steps 1-5
**Approach:**
- Test signup creates token
- Test email sent
- Test verification works
- Test unverified can't login
**Verification:** All tests pass
**Files:** tests/test_auth.py

Example 2: Writing Task

Task: "Write annual report for engineering team"

## Plan

### Step 1: [ANALYSIS] Gather metrics
**Dependencies:** None
**Approach:**
- Pull deployment frequency from CI/CD
- Get incident data from monitoring
- Collect code review stats
**Verification:** Have all data points needed
**Files:** data/raw_metrics.json

### Step 2: [DRAFTING] Write executive summary
**Dependencies:** Step 1
**Approach:**
- Highlight key achievements
- Include year-over-year comparison
- Keep to 1 page
**Verification:** Covers all major points
**Files:** report/executive_summary.md

### Step 3: [DRAFTING] Write detailed sections
**Dependencies:** Step 2 (for tone consistency)
**Approach:**
- Infrastructure improvements
- Team growth and hiring
- Process optimizations
- Major projects delivered
**Verification:** Each section has data + narrative
**Files:** report/sections/*.md

### Step 4: [INTEGRATION] Compile and format
**Dependencies:** Steps 2, 3
**Approach:**
- Combine sections
- Add table of contents
- Apply company template
**Verification:** Document is complete and formatted
**Files:** report/annual_report_2024.md

### Step 5: [REVIEW] Self-review and polish
**Dependencies:** Step 4
**Approach:**
- Check for typos
- Verify all numbers accurate
- Ensure flow between sections
**Verification:** Ready for stakeholder review
**Files:** [same]

Example 3: Data Analysis

Task: "Analyze customer churn for Q4"

## Plan

### Step 1: [DATA] Extract customer data
**Dependencies:** None
**Approach:**
- Query database for customers active in Q3
- Get their Q4 activity status
- Include subscription tier, signup date
**Verification:** Dataset has expected row count
**Files:** data/customers_q4.csv

### Step 2: [ANALYSIS] Calculate churn rate
**Dependencies:** Step 1
**Approach:**
- Define churn: no activity in Q4
- Calculate overall rate
- Break down by tier, acquisition channel
**Verification:** Numbers match expectations
**Files:** analysis/churn_basic.py

### Step 3: [ANALYSIS] Identify patterns
**Dependencies:** Step 2
**Approach:**
- Time-series analysis of churn
- Correlation with features
- Segment analysis
**Verification:** Charts show clear patterns
**Files:** analysis/churn_patterns.ipynb

### Step 4: [SYNTHESIS] Create recommendations
**Dependencies:** Step 3
**Approach:**
- Based on patterns, propose interventions
- Prioritize by impact/effort
- Include success metrics
**Verification:** Recommendations are actionable
**Files:** report/recommendations.md

### Step 5: [DELIVERY] Create presentation
**Dependencies:** Steps 2, 3, 4
**Approach:**
- Executive summary slide
- Key findings with charts
- Recommendations with impact
**Verification:** Stakeholders can act on it
**Files:** presentation/churn_analysis_q4.pptx

Best Practices

1. Make Steps Concrete

❌ Vague: "Set up authentication" ✅ Concrete: "Create users table with bcrypt password hashing"

2. Define Clear Verification

Every step needs a pass/fail check:

  • "Tests pass"
  • "Database table exists with correct schema"
  • "Can send test email successfully"
  • "Response time < 200ms"

3. Keep Steps Small

Break large steps into smaller ones:

❌ "Build entire API" ✅

  • "Create database schema"
  • "Implement GET endpoint"
  • "Implement POST endpoint"
  • "Add validation"
  • "Write tests"

4. Mark Dependencies Explicitly

This prevents trying to build on foundations that don't exist.

5. Plan for Failure

Include what to do if a step fails:

  • Rollback procedure
  • Alternative approaches
  • When to stop and reassess

Comparison: Direct vs Plan-First

Task: "Build a REST API for todos"

Direct ApproachPlan-First Approach
Start coding endpointsPlan database schema first
Discover need auth laterInclude auth in initial plan
Maybe forget validationExplicit validation step
Inconsistent error handlingDefine error format upfront
Tests added at endTest steps interleaved

When Plans Change

Mid-execution discovery: "Actually we need OAuth, not simple auth"

  1. Stop current execution
  2. Reassess remaining plan
  3. Add OAuth-specific steps
  4. Update dependencies
  5. Resume with revised plan

Don't wing it. Update the plan.

Integration with Other Skills

Plan First + Team Code:

  1. Generate master plan
  2. Delegate parallelizable steps to agents
  3. Execute dependency-aware

Plan First + Self-Critique:

  1. Generate plan
  2. Critique the plan ("Am I missing anything?")
  3. Revise
  4. Execute

Quick Start Template

## Task
[Your task here]

## Plan

### Step 1: [VERB] Description
**Dependencies:** None
**Approach:**
- 
- 
**Verification:** 
**Files:** 

### Step 2: [VERB] Description
**Dependencies:** Step 1
**Approach:**
- 
- 
**Verification:** 
**Files:** 

[Continue...]

## Risk Analysis
**Blockers:** 
**Mitigation:** 

## Rollback
**If fails:** 

References

  • Research: "Plan-and-Solve Prompting: Improving Zero-Shot Chain-of-Thought Reasoning by Large Language Models" (Wang et al., 2023)
  • Related: Chain-of-Thought (CoT), Tree-of-Thought (ToT), ReAct
  • See references/examples.md for more detailed examples

平台分布

OpenClaw

70.88%
按下载量换算724

安全审计

VirusTotal

通过

ClawScan

通过

Static analysis

通过

权限和风险

需要联网

该 Skill 可能需要联网访问来源站点、仓库或外部 API;具体网络访问范围需要结合源码和 README 复核。

安装前确认

本站仅展示第三方公开信息,不托管安装包,不提供自动安装或运行环境。安装前应自行审查源码、依赖和命令行为。目前只有一个公开来源,建议同时核对源仓库。

继续浏览同类 Skills