Token导航 LogoToken导航

Agent Skill · 研究检索

pa-onboarding(PA 入职)

pa-onboarding 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。

pa-onboarding 是研究检索类 Agent Skill,可用宿主为 OpenClaw,通过ClawHub获取,权限为接触敏感数据;许可证为 MIT-0,第三方扫描有提醒项,按安装量在研究检索类的 73,675 个 Skill 里排第 27,273。

总安装 3,420周安装 137GitHub Stars 公开资料未说明下载 1,107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
1.1.1

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:pa-onboarding(PA 入职)
来源仓库:https://github.com/netanel-abergel/pa-onboarding
安装命令:
openclaw skills install pa-onboarding
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 pa-onboarding

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。

OpenClaw 新 PA 分步入门指南,涵盖账户配置与技能绑定。适合首次使用者快速上手代理管理与任务分配。提供图文指引与常见问题解答,降低学习成本。部分步骤需人工干预,如邮箱验证或权限审批。建议保存操作指南以备后续查阅。pa-onboarding 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 pa-onboarding;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

    它需要接触敏感数据,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

pa-onboarding 怎么安装?

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

pa-onboarding 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。

pa-onboarding 需要什么权限?

权限级别是接触敏感数据。该 Skill 可能接触密钥、Token、环境变量或敏感配置,应进入高风险复核队列,默认不自动发布。

pa-onboarding 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。

pa-onboarding 用的人多吗?

公开数据是累计安装 3,420 次,近一周安装 137 次,下载 1,107 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。

pa-onboarding 安全吗?

第三方扫描有提醒项,说明存在需要人工确认的行为。安装前请打开源码看清它读了什么、联网到哪里。

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。

先想清楚的地方

  • 需要接触敏感数据,装到常用环境前要先评估影响面。
  • 第三方扫描存在提醒项或未通过项,需要人工逐条确认。
  • 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。

不太适合

  • 环境里有大量密钥、凭证,不方便给第三方 Skill 读取的人。
  • 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
pa-onboarding

PA 入职 · 本页

3,420—接触敏感数据MIT-0有提醒

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
pa-onboarding
description
Step-by-step onboarding guide for setting up a new AI Personal Assistant on OpenClaw. Use when: a new PA is being created, someone asks how to set up an agent, or guiding a user through the full setup process from account creation to first response.

PA Onboarding Skill

Minimum Model

Any model. This is a procedural guide — follow steps in order.


Rules

  • Give one step at a time. Do not dump the full guide upfront.
  • Confirm each step before moving on. Do not assume it worked.
  • Never say something is done unless you verified it.
  • Do not start integrations (calendar, monday) before the agent is responding to messages.

If owner says "what's next?" → give only the single next step. If owner says "I'm stuck on step X" → troubleshoot step X, do not move on. If owner asks "can you do it for me?" → reply: "I can guide you — most steps require your action (QR scan, calendar share, etc.)" If owner asks "how long will this take?" → reply: "Phase 1 takes 20–30 min. Integrations another 15–20 min."


Owner Interaction Signals (Teach This Early)

The agent must learn these signals from day one:

SignalMeaningAgent action
Any task requestOwner delegated a taskReact 👍 immediately, ✅ when done
👍 from ownerGood jobLog positive feedback
👄 from ownerPoor resultFix and log the lesson

Critical rules to cover during onboarding:

  • 👍 before starting any task, ✅ when complete — in every DM with the owner, no exceptions
  • Never ask the owner "did you mean X?" if the answer is inferable — execute and let them correct
  • When owner asks to check on someone: contact that person, then report back to the owner what they said. Never ask the owner instead of the person.
  • Never include the owner's name or internal framing inside messages to third parties (PAs, contacts, vendors)

Phase 1 — Account & Agent (Do This First)

Step 1: Create Agent Account

  • Go to the agent signup page (for monday.com orgs: monday.com/agents-signup).
  • Sign in with organization SSO.
  • Create the agent and give it a name.

Step 2: Get a Phone Number

  • The agent needs a dedicated phone number for WhatsApp.
  • Use Airalo eSIM (or similar) — must support SMS, not data-only.
  • Cost: ~$5–15/month.

Step 3: Install WhatsApp Business

  • Download WhatsApp Business (not regular WhatsApp).
  • Register the new phone number.
  • Complete SMS verification.

Step 4: Connect WhatsApp to Agent

  1. Open OpenClaw platform → Agent Settings → Channels → WhatsApp.
  2. Click Connect → scan QR code with WhatsApp Business app.
  3. Wait 30 seconds for status to show "Connected and listening."

Step 5: Verify the Agent Responds

  • Send a test message to the agent's number.
  • Confirm you get a reply within 30 seconds.
  • If no reply: Stop here. Follow the whatsapp-diagnostics skill before continuing.

✅ Phase 1 complete when the agent responds to messages.


Phase 2 — Integrations

Only start Phase 2 after Phase 1 is fully working.

Step 6: Connect Google Calendar

  • Owner shares their calendar with the agent email.
  • Agent runs:
  gog auth add owner@company.com --services calendar
  • Test that the agent can create events.
  • Full guide: see calendar-setup skill.

Step 7: Connect monday.com

  1. Create a monday.com account for the agent (use org signup URL).
  2. Generate API token: avatar → Developers → My Access Tokens → Copy.
  3. Save the token:
   echo "YOUR_TOKEN" > ~/.credentials/monday-api-token.txt
  1. (Optional) Set up monday MCP for natural language operations.
  2. Full guide: see monday-workspace skill.

Step 8: Set Up Email Access (Optional)

# Authenticate gog with Gmail and other services
gog auth add owner@company.com --services gmail,drive,contacts

# Test it works
GOG_ACCOUNT=owner@company.com gog gmail search 'is:unread' --max 5

Full guide: see openclaw-email-orientation skill.


Phase 3 — Configuration

Step 9: Configure SOUL.md

Key things to include:

  • Owner's name and communication style.
  • Work hours and timezone.
  • What to act on autonomously vs. what requires permission.
  • Topics to proactively monitor.

Step 10: Schedule Morning Briefing (Optional)

  • Cron job at 07:30 owner's timezone, Monday–Friday.
  • Sends: meetings, urgent emails, open tasks.
  • Full guide: see owner-briefing skill.

Step 11: Add to PA Network Directory

  • Update data/pa-directory.json with new PA details.
  • Announce in the PA coordination group.

Common Issues

IssueLikely CauseFix
Agent doesn't respondWhatsApp not properly linkedRe-scan QR code
Messages count = 0Gateway ingest issueRun openclaw gateway restart
Calendar connected but read-onlyWrong share permissionOwner re-shares with "Make changes"
Billing errorAPI key out of creditsTop up or switch model
eSIM not activatingData-only planGet SMS-capable plan

Onboarding Checklist

[ ] Agent account created
[ ] Phone number acquired
[ ] WhatsApp Business installed and registered
[ ] WhatsApp connected to agent
[ ] Agent responds to test message       ← must complete before Phase 2
[ ] Google Calendar connected (write access)
[ ] monday.com account created and token saved
[ ] SOUL.md configured
[ ] Morning briefing scheduled (optional)
[ ] Added to PA network directory
[ ] Announced in PA coordination group

Cost Tips

  • Cheap: Onboarding is procedural — any small model works.
  • Avoid: Do not attempt all phases simultaneously — go in order.
  • Small model OK: Use a medium model only if SOUL.md customization needs nuanced tone matching.

Lessons from Production (2026-04-02)

Lessons learned from running PAs in production. Future agents setting up a new PA should apply these from day 1.

  1. Skill count matters — the sweet spot is 28–32 active skills. Start lean and add a new skill only when there's a clear, recurring trigger phrase that no existing skill covers. Above 40 skills, routing breaks down and the agent starts misfiring. Less is more.
  1. HOT.md — create this file from day 1. Max 20 lines. It's not documentation — it's the short list of rules the agent keeps breaking. If you find yourself correcting the same behavior twice, it belongs in HOT.md. Review it at the start of every session.
  1. SOUL.md vs Skills — universal behavioral rules (temperature selection, proactivity level, tone, response length) go in SOUL.md and apply everywhere. Skills are for triggered, domain-specific workflows only. Don't put behavioral rules inside a skill — they won't apply globally.
  1. Diagnostics are appendices — never create a standalone diagnostics skill (e.g. whatsapp-diagnostics). Instead, add a troubleshooting section at the bottom of the parent skill (e.g. whatsapp). Standalone diagnostics skills fragment routing and add noise to the skill library.
  1. DEPRECATED.md discipline — when you merge or retire a skill, always write a tombstone file explaining why. Include: what it merged into, the date, the reason, and the lesson. Future agents (and humans) will thank you. Without tombstones, retired skills become ghost entries that confuse routing.
  1. pa-skills repo — before building a skill from scratch, check https://github.com/netanel-abergel/pa-skills for a curated starting library. It reflects all the above lessons and provides production-ready skill templates for the most common PA workflows.

平台分布

OpenClaw

89.03%
按下载量换算986

安全审计

VirusTotal

可疑

ClawScan

可疑

Static analysis

通过

权限和风险

接触敏感数据

该 Skill 可能接触密钥、Token、环境变量或敏感配置,应进入高风险复核队列,默认不自动发布。

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。来源安全扫描存在提醒或未通过项,安装前请自行复核。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