HANDS-ON WORKSHOP实战工作坊 SALESFORCE TOWER · TONIGHT悉尼

Build Your
AI Sales
Engine

让 AI 接管
你的 LinkedIn

Watch it work once. Then make it yours. 先看它跑一次，再把它变成你自己的。

Site password · Aaron网站密码 · Aaron 0435 525 634

WiFiWiFi kandji guest Password密码 busybees

Open on your laptop在你的电脑上打开 events.openonion.ai/workshop.html QR to open the workshop deck

01 · ROOM CHECK01 · 现场调查 LISTEN FIRST先听大家

Know the Room First

先认识一下彼此

01What do you do?你目前从事什么工作？

02How long have you used Codex?你使用 Codex 多久了？

03What do you want to learn today?今天最希望学会什么？

02 · TODAY02 · 今天 TWO WORKFLOWS两条工作流

What You'll Learn Today

今天你会学到什么

Flow 1 · Engagement工作流 1 · 互动Warm your network — like and comment.点赞、评论，焐热已有人脉。

Flow 2 · Outreach工作流 2 · 触达新人Find your ICP · write in your voice.找到 ICP · 用你的语气写。

Write good prompts写好 promptExamples over rules · ten drafts, pick one.例子胜过规则 · 写十条挑一条。

Codex & your browserCodex 和你的浏览器Stealth browser · everything kept as skills.隐身浏览器 · 一切留成 Skill。

Two workflows — skills you keep.两条工作流——Skill 归你。

03 · LIVE DEMO03 · 现场演示 REAL ACCOUNT · COMPLETE FLOW真实账号 · 完整流程

Complete a LinkedIn Engage

完成一次完整的 LinkedIn Engage

STEP 01Browse the feed浏览动态

STEP 02Choose a post选择帖子

STEP 03Pick a reaction选择 Reaction

STEP 04Write & publish生成并发布评论

This is live: the AI reads first, then decides how to engage.不是预录：让大家看到 AI 先读内容，再决定如何互动。

04 · WHY IT WORKS04 · 为什么有用 WARM THEM UP先预热

Why Engage First

为什么先互动，再开口

They're already here人已经在了Your ICP is already in your network.ICP 就在你的人脉里。

Warm, then ask先热，再开口Show up before you ask.先出现，再开口。

The algorithm notices算法看得见Active account → more reach.账号活跃 → 触达更多。

It comes back会还回来They come back to yours.他们也会回到你的帖子。

Every like is a deposit — paid back as reach.每个赞都是存款——以触达偿还。

05 · MEET CODEX05 · 认识 CODEX WHERE YOU TYPE你在哪打字

This Is Codex

这就是 Codex —— 你打字的地方

CodexCodexPart of ChatGPT — one box at the bottom.属于 ChatGPT——底部一个输入框。

ConnectOnionConnectOnionThe stealth browser. “co” in commands.隐身浏览器。命令里叫 “co”。

You你Paste, Enter, watch.粘贴、回车、看着。

Everything happens in that one box.今天的一切都在这一个框里。

06 · CODEX MOVES06 · CODEX 技巧 TOGGLE · FILES · APPROVE切面板 · 看文件 · 批准

A Few Codex Moves

用 Codex 顺手的几招

Toggle the panel切换右侧面板Terminal ⌘T · Browser ⌘P · Files · Side chat ⌥⌘STerminal ⌘T · Browser ⌘P · Files · Side chat ⌥⌘S

Open Files打开 FilesSkills and screenshots appear as it works.skill 和截图边跑边冒出来。

One input box就一个输入框Paste a skill, press Enter.粘贴 skill，按回车。

Approve when asked它问就批准Click Approve — your one job.点 Approve——你唯一的活。

Watch, and click Approve.看着，点 Approve。

07 · SKILL = PROMPT + TOOLS07 · SKILL ＝ PROMPT + 工具 ONE FILE一份文件

Skill = Prompt + Tools

Skill ＝ Prompt + 工具

Skill＝ Prompt＋ Tools

Skill ＝ Prompt ＋工具

Built-in: browser, terminal, files · plug-in: CLI, API.自带：浏览器、终端、文件 · 外接：CLI、API。

---
name: skill-name
description: when to use it
---

# Steps
1. do this
2. then this
3. stop and confirm

08 · YOUR FIRST SKILL08 · 第一个 SKILL COPY, PASTE, RUN复制、粘贴、运行

A Real Skill: Sydney Weather

一个真实的 Skill：悉尼天气

---
name: sydney-weather
description: Check today's weather in Sydney and say whether I need an umbrella.
---

# Sydney Weather

1. Fetch https://wttr.in/Sydney
   (use your browser tool if you have one).
2. Read the current temperature and the chance of rain.
3. Answer in one line: temperature + umbrella or not.

Read it. Copy it. Run it.读一读，复制，跑起来。

A complete Skill. Paste into Codex.完整的 Skill。粘进 Codex。

09 · CREATE A PROJECT09 · 新建项目 A PROJECT IS A FOLDER项目就是一个文件夹

Create a Project

新建一个项目

Create→ Project

点 Create → 选 Project → 起个名字

A project is just a folder.项目就是一个文件夹。

10 · WHY CONNECT10 · 为什么用 CONNECT LINKEDIN

LinkedIn Detects the Browser

LinkedIn 会检测浏览器

Codex BrowserCodex 浏览器

Detected被检测

ConnectOnion BrowserConnectOnion 浏览器

Undetected检测不到

The browser behind our A$2,000 client automations.A$2,000 级客户方案用的就是它。

11 · STEALTH BROWSER11 · 隐身浏览器 COPY → PASTE → DONE复制 → 粘贴 → 完成

Install the Stealth Browser

一键装好隐身浏览器

---
name: install-connectonion
description: Install ConnectOnion and configure everything on Windows, macOS, or Linux — find Python, install the package, scaffold a project, set up API keys (managed or own), install a browser if needed — then verify with a real run and hand the user a plain-language account summary. Use when the user says "install connectonion", "set up connectonion", "get me started with co", or a co command fails because nothing is configured yet.
tools:
  - Bash(python *)
  - Bash(python3 *)
  - Bash(py *)
  - Bash(pip *)
  - Bash(pip3 *)
  - Bash(co *)
  - Bash(patchright *)
  - shell
  - read_file
  - write
  - edit
---

# Install ConnectOnion Skill

Take a machine from nothing to a working, configured agent — on **whatever operating
system the person is using** — then tell them, in plain words, exactly what they have.

**Assume the person running this may not be technical.** They may not know what a
"virtual environment", a "PATH", or an "API key" is, and that's fine — your job is to do
the technical work *for* them and keep them informed in friendly, jargon-free language.
Never dump a raw stack trace at them; translate every problem into "here's what happened
and here's what I'm doing about it."

## How to run this skill

1. **Work on the user's real OS.** Windows, macOS, and Linux differ in command names and
   shells. Do Step 0 first and use the right commands for that platform — don't assume
   `python3`, `grep`, or `bash` exist everywhere.
2. **Verify every step before moving on.** Never assume a command worked — check its
   output. Each step has a check; run it.
3. **Auto-correct instead of giving up.** When a step fails, do NOT stop and paste the
   error. Diagnose the cause, apply the known fix (the *Recovery* box under each step),
   and retry. Give each fix **up to 2 attempts** before escalating.
4. **Only ask the human for things only a human can do** — logging in through a browser,
   adding money, choosing a project name. Ask in one plain sentence.
5. **Keep them in the loop.** A short friendly line before each phase ("Installing now —
   about 30 seconds…") beats silence.
6. **Never print or store secrets.** Confirm a key is *present*, never show its value.

## Step 0: Know the machine

Figure out the platform and which commands to use. **Everything below writes `PY` for
"the Python command that works on this machine"** — resolve it once here, then reuse it.

- **Detect the OS.** If you have a shell, `uname` (prints `Darwin`/`Linux`) vs. failing =
  Windows is a quick tell; otherwise use whatever platform signal you have.
- **Find Python (needs 3.10+).** Try these in order and keep the first that prints 3.10 or
  newer — that is `PY`:

  ```bash
  python3 --version      # macOS / Linux usually
  python --version       # Windows usually, sometimes macOS/Linux
  py -3 --version        # Windows Python launcher
  ```

- **Pick the right shell.** On macOS/Linux the `bash` tool is fine. **On Windows the
  `bash` tool does not work** (it's Unix-only and returns an error) — use the
  cross-platform `shell` tool instead, and don't rely on `grep`/`cat`/`ls`. Prefer
  `co …` commands and `PY -c "…"` snippets, which behave the same everywhere.

> **Recovery**
> - None of `python3` / `python` / `py -3` is found, or all are < 3.10 → Python isn't
>   installed (or is too old). Tell the person plainly and give the path for *their* OS:
>   - **Windows:** install from python.org (tick "Add python.exe to PATH"), or
>     `winget install Python.Python.3.12`. Afterward `py -3` works.
>   - **macOS:** install from python.org, or `brew install python@3.12`.
>   - **Linux:** use the distro package manager (e.g. `sudo apt install python3 python3-venv python3-pip`).
>   Don't install system Python yourself unless they ask.

## Step 1: Isolate with a virtual environment (recommended)

Keeps this install tidy and avoids "externally-managed-environment" errors:

```bash
# macOS / Linux
PY -m venv .venv && source .venv/bin/activate

# Windows (PowerShell)
py -3 -m venv .venv ; .venv\Scripts\Activate.ps1
# Windows (cmd)
py -3 -m venv .venv & .venv\Scripts\activate.bat
```

After activating, `python`/`pip` point inside `.venv` — from here `PY` can just be
`python`. If venv creation fails, it's optional: skip it, continue globally, and note
that to the user.

## Step 2: Install the package

Use `PY -m pip` (works even when a bare `pip` isn't on PATH):

```bash
PY -m pip install connectonion       # add -U to upgrade an existing install
```

Verify before continuing:

```bash
co --version    # prints the installed version
co doctor       # cross-platform health check: Python, config, keys, browser
```

> **Recovery**
> - `co: command not found` after a successful install → the scripts folder isn't on
>   PATH. Confirm the package works with `PY -m connectonion.cli.main --version`; if that
>   prints a version, either use `PY -m connectonion.cli.main …` for the rest, or reinstall
>   with `pipx install connectonion` (which puts `co` on PATH).
> - "externally-managed-environment" / permission error → go back to Step 1 and use the
>   venv (exactly what it prevents), then retry.

## Step 3: Create or initialize a project

A project is just a folder with a `.env` (its keys) and a `.co/` folder (its config +
identity). The CLI is cross-platform here — the same commands work on every OS.

**New project** — **always pass `--yes`** so it doesn't stop to ask questions:

```bash
co create my-agent --yes                            # minimal (default)
co create my-coder --template coder --yes           # bash + file editing
co create my-bot --template browser --yes           # browser automation (see Step 6)
co create my-researcher --template web-research --yes
```

Other templates: `hosted-browser`, `co-ai`, `custom` (with `--description "..."`).

**Existing directory** — add config in place: `co init --yes` (config only) or
`co init --template minimal --yes` (full starter). The first run on a machine also
creates the global `~/.co/` config (keypair, agent address, agent email) automatically.

Confirm it worked with a cross-platform check (don't use `ls` — it's not on Windows):

```bash
co doctor       # its Config/Keys lines show ✓ when .co/ and keys exist
# or, portably:
PY -c "import os; print('OK' if os.path.exists('.env') and os.path.isdir('.co') else 'MISSING')"
```

> **Recovery**
> - "Directory not empty" warning → the folder already has files. Add `--force` only after
>   the person confirms it's safe to write here.
> - The command hangs → it's waiting on an interactive prompt; you forgot `--yes`. Re-run
>   with `--yes`.
> - `.env` / `.co/` missing afterward → re-run and read the output before continuing.

## Step 4: Configure model access (pick ONE path)

An agent can't think without a model. Two paths — pick by what the person has.

**Path A — managed keys (recommended; no provider account, comes with free credit):**

```bash
co auth
```

Logs them in and saves `OPENONION_API_KEY` to `~/.co/keys.env` (and the project `.env`).
Agents then use `co/*` models — `co/o4-mini`, `co/gpt-4o`, `co/claude-sonnet-4-5`,
`co/gemini-2.5-pro` — with nothing else to configure. `co auth` opens a browser login, so
hand off: *"Please finish the login in the browser window that opened — I'll continue
once you're back."*

**Path B — their own provider key:** add it to the project `.env` with the `write`/`edit`
tool (never echo the value back afterward):

```
OPENAI_API_KEY=sk-...        # for gpt-* models
ANTHROPIC_API_KEY=sk-ant-... # for claude-* models
GEMINI_API_KEY=...           # for gemini-* models
```

Model name picks the provider: `gpt-*` → OpenAI, `claude-*` → Anthropic, `gemini-*` →
Google, `co/*` → managed keys.

Verify (cross-platform — `co doctor` reports where it found the key, no `grep` needed):

```bash
co status       # Path A: shows account, balance, and free credit
co doctor       # either path: the "API Key" line shows ✓ Found in .env / keys.env
```

> **Recovery**
> - `co auth` login didn't complete → safe to re-run; it refreshes the token in place.
> - "No API key found" → the key didn't land. Re-run `co auth` (Path A) or re-check the
>   variable name spelling in `.env` (Path B).

## Step 5: Verify with a real run

Prove the whole chain works with a tiny 2-line agent. **Match the model to the path from
Step 4** — the default model is a managed `co/*` one, so a Path-B setup must name its own
provider's model:

```bash
# Path A (managed keys — default model):
PY -c "from connectonion import Agent; print(Agent('smoke-test', max_iterations=2).input('Reply with exactly: OK'))"

# Path B (own key — name a model your key serves, e.g. OpenAI):
PY -c "from connectonion import Agent; print(Agent('smoke-test', model='gpt-4o-mini', max_iterations=2).input('Reply with exactly: OK'))"
```

Expected: a reply containing `OK`.

> **Recovery — read the error and act, don't just report it**
> - `InsufficientCreditsError` → the managed account is out of credit. A *money* thing only
>   the person can fix: show the address and balance from `co status`, point them to top up
>   (https://o.openonion.ai/purchase) or Discord (https://discord.gg/4xfD9k8AUF). Not an
>   install failure.
> - Auth / `401` → the key from Step 4 didn't land. Re-run `co auth` (A) or fix `.env` (B),
>   then retry once.
> - `ModuleNotFoundError: connectonion` → wrong Python/venv is active. Re-activate the venv
>   from Step 1 and retry.

## Step 6: Browser setup — ONLY if they need web automation

Skip this entirely unless the person chose the `browser` / `hosted-browser` template or
wants the agent to drive a web browser. **This is a real gotcha:** `pip install
connectonion` installs the browser *library* (`patchright`) but **NOT a browser to
drive**. Without one, browser commands fail.

The agent looks for a **real Google Chrome** first (auto-detected on Windows, macOS, and
Linux). If Chrome isn't installed, install Patchright's own browser:

```bash
PY -m patchright install chrome    # downloads a Chromium the agent can drive
```

Verify:

```bash
co doctor       # the Browser line reports ok / broken / missing
```

> **Recovery**
> - `co doctor` browser line says **missing** → run `PY -m patchright install chrome`.
> - Says **broken** (stealth driver) → `PY -m pip install --force-reinstall --no-cache-dir patchright`, then re-run `patchright install chrome`.
> - Still failing → check `~/.co/browser.log`; the person may simply install desktop Google
>   Chrome, which the agent will pick up automatically.

## Step 7: Hand the user a plain-language summary

The payoff. Run `co status` and turn it into a short, friendly summary — **don't paste the
raw panel**, translate it.

```bash
co status
```

Map the `co status` fields to plain language:

| `co status` field | Say it like this |
|--------------------|------------------|
| Credits            | 🎁 **Free credit from ConnectOnion** — money they gave you to start |
| Balance            | 💰 **Money available right now** to run your agent |
| Total Spent        | what's been used so far |
| Email              | 📧 **Your agent's own email address** (it can send/receive mail) |
| Agent ID / Address | your agent's unique ID |

Then present something like this (fill in the real values):

```
✅ You're all set up! Here's your ConnectOnion account:

  🎁  Free credit ConnectOnion gave you:  $X.XX
  💰  Money available to use now:          $X.XX
  📧  Your agent's email address:          you@mail.openonion.ai
  🤖  Model access:                        managed (co/* models) — nothing else needed
  📦  Installed:                           connectonion vX.Y.Z  (Python 3.12 on Windows)
  📁  Your project:                        ./my-agent  (minimal template)

  You can start using your agent now. If the balance ever runs low, you can add
  more at https://o.openonion.ai/purchase — the free credit is enough to get going.
```

Adjust for the situation:
- **Balance $0 / low** → say it gently: "Your free credit is used up, so you'll need to add
  a little to keep running the agent — here's where: …". No alarm, no jargon.
- **Path B (own keys)** → replace the money lines with "Model access: your own <provider>
  key"; skip balance (their provider bills them directly).
- **Browser installed** → add "🌐 Browser: ready".
- **Integrations connected** → add "📬 Connected: Gmail" etc.

Keep it to the handful of lines that matter to *this* person. Warmth over completeness.

## Notes

- **Cross-platform:** the `co` CLI runs on Windows, macOS, and Linux. The gotchas are (1)
  the `bash` *tool* is Unix-only — use `shell` on Windows; (2) `python3` may be `python`
  or `py -3` on Windows (that's why Step 0 resolves `PY`); (3) the browser binary is never
  auto-installed (Step 6).
- `--yes` on `co create` / `co init` is mandatory for unattended runs, or the command
  blocks on prompts.
- Never print, log, or commit API keys. Templates gitignore `.env` — keep it that way.
- `co auth` is safe to re-run; it refreshes the token in place.
- A failing smoke test with a green `co doctor` / `co status` is a model-or-credit issue,
  not an install issue — say so plainly.
- All state lives in `~/.co/` (global identity + keys) and the project's `.co/` + `.env`.
  Deleting one project never breaks another.

One copy. One paste.复制一次，粘贴一次。

Codex installs it · emails a check-in when done.Codex 全自动安装 · 装完发签到邮件。

When Codex asks — click Approve.Codex 询问时，点 Approve。

12 · CHECK & HELP12 · 检查 & 答疑 HANDS-ON · ~10 MIN现场动手 · 约 10 分钟

Did It Work?

先确认每台电脑都装好了

GREENCodex says “you’re all set”Codex 显示"全部就绪"

STUCKRaise your hand — we fix it now卡住就举手，现在就修

CHECK-INYour install pinged me安装已给我发信号

Nobody moves on until every screen is green.每台电脑亮绿，才往下走。

13 · MAKE A SKILL13 · 写一个 Skill COPY → RUN复制 → 运行

Your First Browser Skill

一个只做一件事的浏览器 Skill

---
name: linkedin-login-check
description: Open LinkedIn in the ConnectOnion browser and tell me whether I am logged in.
---

# Check My LinkedIn Login

Use the ConnectOnion browser for every step — never any other browser.

## How to drive the ConnectOnion browser

Run these in the shell. This is the browser LinkedIn cannot detect — never use
any other browser, and never use a plain HTTP fetch.

  co browser go_to "<url>"                 # navigate
  co browser get_text                       # read the visible page
  co browser click "<plain description>"    # e.g. "the Like button on the first post"
  co browser scroll                         # load more of a feed
  co browser take_screenshot .tmp/shot.png
  co browser wait_for_manual_login LinkedIn # pause so I can log in once
  # Never use "co browser do" — it starts a second, separate browser agent (unsafe and unpredictable). Use only the explicit commands above.

  co browser help                           # every command, if you need one not listed here


1. Open LinkedIn in the ConnectOnion browser.
2. If it asks me to log in, wait while I log in once — the session is saved after that.
3. Take a screenshot.
4. Tell me in one line: am I logged in — yes or no?

One skill. One job.一个 Skill，一件事。

Paste to save · run it next slide.粘贴保存 · 下一页运行。

14 · RUN IT14 · 运行它 TYPE · ENTER输入 · 回车

Now Run It

现在运行它

$linkedin-login-check

Browser opens — log in once.浏览器打开——登录一次。

15 · Part 115 · 第一部分 MARKETING BASICSMARKETING 基本功

Part 1第一部分

Engagement

互动

Like and comment first — ask later.先点赞、先评论——再开口。

16 · FIRST LIKE16 · 第一个赞 LIKE ONLY只点赞

Your First Like

跟一个真人打个招呼

---
name: linkedin-like
description: Open Aaron's workshop post in the ConnectOnion browser and like it. Just a like — no message, no connection request.
---

# Like Aaron's Post

Use the ConnectOnion browser for every step — never any other browser.

## How to drive the ConnectOnion browser

Run these in the shell. This is the browser LinkedIn cannot detect — never use
any other browser, and never use a plain HTTP fetch.

  co browser go_to "<url>"                 # navigate
  co browser get_text                       # read the visible page
  co browser click "<plain description>"    # e.g. "the Like button on the first post"
  co browser scroll                         # load more of a feed
  co browser take_screenshot .tmp/shot.png
  co browser wait_for_manual_login LinkedIn # pause so I can log in once
  # Never use "co browser do" — it starts a second, separate browser agent (unsafe and unpredictable). Use only the explicit commands above.

  co browser help                           # every command, if you need one not listed here


1. Open this exact post in the ConnectOnion browser:
   https://www.linkedin.com/posts/aaron404-com_today-im-teaching-a-codex-workshop-for-people-activity-7484079776526229504-dVNB
2. Check it opened — it is Aaron's post about teaching a Codex workshop.
   If it did not open, stop and tell me.
3. Like it. That is the whole job — do not comment, do not connect,
   do not message.
4. Screenshot, then tell me in one line: liked, yes or no.

Your first like.你的第一个赞。

The safest action on LinkedIn — one like on my post.LinkedIn 上最安全的动作——给我的帖子点一个赞。

17 · RUN THE LIKE17 · 运行点赞 TYPE · ENTER输入 · 回车

Run the Like

运行这次点赞

$linkedin-like

Enter — it opens my post and likes it.回车——打开我的帖子，点个赞。

18 · IMPROVE IT18 · 改进它 CHANGE ONE LINE改一行就行

Make It Yours

改一行，点赞你行业的帖子

# linkedin-like, step 1 — before:
1. Open Aaron's post and like it.

# after — make it yours:
1. Search LinkedIn for a recent post about
<my industry> and like that one instead.

Change one line · run again · your world.改一行 · 再跑一遍 · 你的领域。

19 · COMMENT WELL19 · 写好评论 DRAFT · CHECK · REVISE起草 · 自检 · 改写

Like, Then Comment

点赞，再写一条好评论

---
name: linkedin-comment
description: Open a LinkedIn post in the ConnectOnion browser, like it, then draft a genuine comment in my voice — checking and rewriting it until it is good. Draft only: show it to me, do not post.
---

# Comment on a LinkedIn Post

Use the ConnectOnion browser for every step — never any other browser.

## How to drive the ConnectOnion browser

  co browser go_to "<url>"
  co browser get_text
  co browser click "<plain description>"
  co browser take_screenshot .tmp/shot.png
  co browser help

  # Never use "co browser do" — it starts a second, separate browser agent (unsafe and unpredictable). Use only the explicit commands above.

1. Open the post I give you (or my feed) in the ConnectOnion browser.
2. Read the post fully. Like it.
3. Draft a comment in my voice, then CHECK your own draft against this bar:
   - Does it react to something specific in THIS post (not generic praise)?
   - Does it add a thought, a question, or experience — not just "Great post!"?
   - Does it sound like a person, not a brand? No "synergies", no emojis-as-filler.
   - One or two sentences. Warm. Never salesy.
4. If the draft fails any check, rewrite it and check again. Loop until it passes.
   Do not show me weak drafts — only bring me one you would be proud to post.
5. Show me the final comment next to the post — but do NOT post it.
6. This is practice: we are writing the comment, not publishing it. Stop after
   showing me. Only post later if I explicitly tell you to.

A comment worth posting.值得发出去的评论。

Drafts · grades itself · rewrites · you approve.起草 · 自检 · 改写 · 你确认。

37 · THE HARD ONES37 · 最难的那种 SAVE THE PAGE · THEN BUILD先存页面 · 再构建

Build a Skill from a Real Page

最复杂的流程：让浏览器照着真实页面把 Skill 建出来

# Example: build a “linkedin-easy-apply” skill

Some flows are too fiddly to hand-write — a job application
with several steps, a resume upload, a Review screen. For
those, don't guess the buttons. Let the builder read the
real page and build the skill from it.

1. Open a job that has “Easy Apply” in the ConnectOnion browser.
2. SAVE THE PAGE FIRST — the builder works from the real form:
     co browser save_page_context easy-apply
3. The builder reads that saved HTML/CSS and finds the actual
   fields — name, email, phone, resume upload, the Next button —
   instead of selectors that break next week.
4. It fills each step and screenshots to verify before moving on.
5. It stops at Review. You approve before it ever submits.

Save the page, then build, then run. That order is the trick:
a skill built from the real page keeps working; a guessed one
breaks the moment LinkedIn changes a field.

For the hard, multi-step ones.给最难的多步流程。

Too complex to hand-write? Built from the page itself.手写搞不定？照着页面替你建。

20 · RUN THE COMMENT20 · 运行评论 TYPE · ENTER输入 · 回车

Run the Comment

运行这条评论

$linkedin-comment

Enter — like, comment, rewrite. Posts only on your yes.回车——点赞、评论、改写。你说 yes 才发。

21 · THE CRAFT21 · 写 SKILL 的手艺 SHOW · DON'T TELL给例子 · 别讲规则

Write a Good Skill

写好 Skill 的四个手艺

Examples > rules例子 > 规则Two real examples beat twenty rules.两个真实例子胜过二十条规则。

Write 10, keep the best写 10 条，留最好的Best-of-ten beats one-shot.十选一赢过一次成型。

First principles第一性原理Define “good” for THIS task, in your words.用你的话定义这件事的"好"。

Get it reviewed让它被评审A second agent critiques the draft.让另一个 agent 挑毛病。

Every skill you copied today was written this way.今天复制的每个 Skill，都是这么写出来的。

22 · PICK THE MODEL22 · 选模型 NEWER = MORE COMPLEX越新 = 越能扛复杂

Which Model to Run It On

同一个 Skill，换模型结果会不一样

Newer handles more越新越能扛10+ steps: 5.6 handles it · 5.5 struggles · 5.4 loses the thread.10 步以上：5.6 扛得住 · 5.5 勉强 · 5.4 跟丢。

Claude goes deepClaude 更能啃硬活Heavier, more complex operations.更能啃重、复杂的活。

Skills are portableSkill 是通用的Same skill, either model — slightly different results.同一个 Skill 两边能跑，结果略有差别。

Run the best直接上最好的4 resets: A$40 → A$120. Run 5.6 for everything.4 次 reset：A$40 → A$120。全程上 5.6。

Don't overthink it — just run 5.6.别纠结——直接上 5.6。

23 · Part 223 · 第二部分 OUTREACH BASICS私信基本功

Part 2第二部分

Find & Message

找人 & 私信

Right people · your voice · you approve every send.找对的人 · 你的语气 · 发送你确认。

24 · FIRST CONNECT24 · 第一个连接 FIND · VERIFY · CONNECT找到 · 核对 · CONNECT

Connect with Aaron

先跟一个真人建立连接——就是我

---
name: linkedin-connect
description: Find Aaron on LinkedIn in the ConnectOnion browser, check it is the right Aaron, and send a connection request with a short note.
---

# Connect with Aaron

Use the ConnectOnion browser for every step — never any other browser.

## How to drive the ConnectOnion browser

  co browser go_to "<url>"
  co browser get_text
  co browser click "<plain description>"
  co browser take_screenshot .tmp/shot.png
  co browser help

  # Never use "co browser do" — it starts a second, separate browser agent (unsafe and unpredictable). Use only the explicit commands above.

1. Open https://www.linkedin.com/in/aaron404-com/ in the ConnectOnion browser.
2. Check this is the right person: the headline should mention
   ConnectOnion or OpenOnion. If it does not match, stop and tell me.
3. Click Connect. If it is hidden behind "More", look there. Choose "Add a note".
4. Write a short note in my voice — I am at Aaron's workshop today, just built my
   first LinkedIn agent, would love to connect. Keep it short; LinkedIn caps it.
5. Send the request. Screenshot, then tell me in one line: connection request sent.

Your first connection.你的第一个连接。

Find me · verify · connect with a note. Then aim it at your ICP.找到我 · 核对 · 带 note 连接。下一步对准你的 ICP。

25 · RUN THE CONNECT25 · 运行连接 TYPE · ENTER输入 · 回车

Run the Connect

运行这次连接

$linkedin-connect

Type · Enter · your note reaches me输入 · 回车 · 你的 note 直达我

26 · FIND PEOPLE26 · 找到目标人 SEARCH LINKEDIN在 LinkedIn 上搜人

Find the Right People

找到你要找的人

---
name: linkedin-find-people
description: Search LinkedIn for people who match my target and list the best ones.
---

# Find People on LinkedIn

Use the ConnectOnion browser for every step — never any other browser.

## How to drive the ConnectOnion browser

Run these in the shell. This is the browser LinkedIn cannot detect — never use any other browser, and never use a plain HTTP fetch.

  co browser go_to "<url>"                 # navigate
  co browser get_text                       # read the visible page
  co browser click "<plain description>"    # e.g. "the Like button on the first post"
  co browser scroll                         # load more of a feed
  co browser take_screenshot .tmp/shot.png
  co browser wait_for_manual_login LinkedIn # pause so I can log in once
  # Never use "co browser do" — it starts a second, separate browser agent (unsafe and unpredictable). Use only the explicit commands above.

  co browser help                           # every command, if you need one not listed here


1. Open LinkedIn People Search in the ConnectOnion browser.
2. Search for my target:
   <role / industry / city — e.g. "support coordinator, NDIS, Sydney">
3. Read the first page (skip ads and "people you may know").
4. Return the top 5: name, headline, profile link.

Who are you looking for?你要找的是谁？

Plain words in → top 5 with links大白话进 → 前 5 人带链接

27 · RUN THE SEARCH27 · 运行找人 TYPE · ENTER输入 · 回车

Run the Search

运行：找人

$linkedin-find-people

Role · industry · city → five people职位 · 行业 · 城市 → 五个真人

28 · WRITE-MESSAGE SKILL28 · 写消息 SKILL THE PROMPT · WITH EXAMPLES提示词 · 含范例

Skill 1 — Write the Message

技能一 —— 写消息

---
name: linkedin-write-message
description: Write a short, human first message to one LinkedIn person, in my voice.
---

# Write My LinkedIn Message

Inputs: their name, headline, one real thing about them; who I am and what I help with.

Rules:
- 2 to 4 sentences. Warm and specific. Never salesy.
- Open with the one real thing about them.
- End with a light, easy question.

Good example:
"Hi Sam — your post on cutting NDIS admin time hit home. I help coordinators automate the boring half. Curious how you're handling intake right now?"

Never write:
"I'd love to connect and explore potential synergies."

Teach it your voice.把你的语气教给它。

One good example beats ten rules一个好例子胜过十条规则

29 · RUN THE WRITER29 · 运行写消息 TYPE · ENTER输入 · 回车

Run the Writer

运行：写消息

$linkedin-write-message

Pick one person · it writes like you挑一个人 · 它写得像你

30 · PICK THE BEST30 · 挑出最好的 10 DRAFTS → PICK 1-2写10条 → 挑1-2条

Ten Drafts, Then the Best One

写十条，挑出最好的

Use my linkedin-write-message skill to write 10 different first-message drafts for <this person>.

Then have separate reviewers score all 10 on four angles:
1. Sounds human, not AI
2. Specific to this person
3. Not salesy
4. Would actually get a reply

Show me only the best 1 or 2 — with one line on why each won.

Let it compete with itself.让它自己跟自己比。

One draft is a guess — ten compete一条是碰运气 — 十条来竞争

31 · SEND-MESSAGE SKILL31 · 发消息 SKILL VERIFY → ONE-SHOT SEND核对 → 一次发出

Skill 2 — Send the Message

技能二 —— 发消息

---
name: linkedin-send-message
description: Send one approved message to one verified LinkedIn person — one shot, after I confirm. Always use the ConnectOnion browser.
---

# Send a LinkedIn Message

Use the ConnectOnion browser for every step — never any other browser.
LinkedIn only shows a Message button once you are connected, so this is also where a connection request happens — with my approved words in the note.

## How to drive the ConnectOnion browser

Run these in the shell. This is the browser LinkedIn cannot detect — never use any other browser, and never use a plain HTTP fetch.

  co browser go_to "<url>"                 # navigate
  co browser get_text                       # read the visible page
  co browser click "<plain description>"    # e.g. "the Like button on the first post"
  co browser scroll                         # load more of a feed
  co browser take_screenshot .tmp/shot.png
  co browser wait_for_manual_login LinkedIn # pause so I can log in once
  # Never use "co browser do" — it starts a second, separate browser agent (unsafe and unpredictable). Use only the explicit commands above.

  co browser help                           # every command, if you need one not listed here


1. Open the chosen person's profile in the ConnectOnion browser.
   Verify the name and headline match.
2. Take whichever path the profile offers:
   - If there is a Message button (we are connected), open it.
   - If there is not, click Connect and choose "Add a note" — the note carries
     my message, and it reaches them without waiting for them to accept.
   - If Connect is hidden behind "More", look there.
3. Paste the message I approved — do not change a word.
4. Show me a screenshot before sending.
5. Send once. Never retry the send button. Tell me which path you took.

The AI drafts. You hit send.AI 起草，你按发送。

Verify the person · one-shot the send核对对的人 · 一次发出

32 · THE BUILDER32 · 构建者 SKILLS WRITE SKILLSSKILL 生 SKILL

This Skill Wrote Itself

这个 Skill 是 AI 写的

The builder构建者A Skill that writes Skills from a real page.一个从真实页面生成 Skill 的 Skill。

Take it home — it builds the next one for you.带回家——下一个 Skill 让它替你写。

32 · RUN THE SENDER32 · 运行发消息 TYPE · ENTER输入 · 回车

Run the Sender

运行：发消息

$linkedin-send-message

Nothing sends until you say yes你说可以，才会发出

33 · PUT IT TOGETHER33 · 串起来 SKILLS CALLING SKILLSSKILL 调用 SKILL

One Command, The Whole Flow

一条命令，跑完整个流程

---
name: linkedin-daily
description: My whole LinkedIn routine in one go — find the right people, write to each in my voice, and send only what I approve. Uses the ConnectOnion browser.
---

# My LinkedIn Routine

Use the ConnectOnion browser for every step — never any other browser.

## How to drive the ConnectOnion browser

  co browser go_to "<url>"                 # navigate
  co browser get_text                       # read the visible page
  co browser click "<plain description>"    # click by description
  co browser take_screenshot .tmp/shot.png
  co browser help                           # every command

  # Never use "co browser do" — it starts a second, separate browser agent (unsafe and unpredictable). Use only the explicit commands above.

Ask me who I am targeting today, then:

1. Use my linkedin-find-people skill — get 5 matches.
2. Show me the 5. Let me drop any I do not want.
3. For each one I keep:
   a. Use my linkedin-write-message skill to draft a message.
   b. Show me the draft next to their profile. Wait for my yes.
   c. On yes, use my linkedin-send-message skill. On no, skip them.
4. Report: who you found, what went out, who I skipped.

Rules:
- Never send anything I have not said yes to.
- One message per person. Never retry a send.
- If anything looks off — wrong person, odd draft — stop and ask me.

Skills calling skills.Skill 调用 Skill。

Three skills, one routine. You still approve every send.三个 Skill，一条流程。每次发送仍由你批准。

34 · AUTOMATION34 · 自动运行 AI PREPARES · PEOPLE DECIDEAI 准备，人决定

Automate Only After It Is Stable

稳定以后，再让它自动运行

Prepare Overnight夜间准备

Content · assets · tomorrow’s queue生成内容 · 整理素材 · 排好队列

Review in Daylight白天确认

Check outputs · people confirm sends查结果 · 发送由人确认

Undetected ≠ safe. Low volume · approve every send.隐身 ≠ 安全。量要小 · 每次发送人工确认。

Production service: monitoring · upgrades · maintenance生产服务：监控 · 升级 · 维护

36 · THE ONE LIMIT36 · 关键边界 REMOTE IS NOT THE HOSTREMOTE 不是运行主机

The Computer Still Does the Work

手机负责指挥，真正干活的仍是电脑

Phone sends手机发送

Prompts · approvals · decisions指令 · 批准 · 决策

Computer provides电脑提供

Files · credentials · browser · local tools文件 · 账号 · 浏览器 · 本地工具

Host sleeps → Remote pauses. Mac mini optional.电脑休眠，Remote 暂停。Mac mini 可选。

39 · BEFORE YOU GO39 · 走之前 2 MINUTES · THEN PIZZA两分钟 · 然后披萨

Feedback & Networking

反馈 & 交流

Feedback反馈Two minutes, right now — or the WhatsApp group.现在两分钟——当面说，或发群里。

Networking交流Pizza is out — meet the people beside you.披萨已就位——认识身边的人。

One intro一次引荐One intro before you leave.走之前认识一个人。

The room is the point. Stay as long as you like.重点是这间屋子。想待多久都行。

40 · WRAP UP40 · 收尾 RECAP · COMMUNITY · NEXT STEP回顾 · 社区 · 下一步

Take the Workflow Home

把今天的流程带回家

Today今天Browser → Skill → your own workflow浏览器 → Skill → 你自己的流程

At home回家One login: co auth microsoft登录一次：co auth microsoft

Docs文档docs.connectonion.com

These slides这份课件Download this deck — one file, works offline下载这份课件——一个文件，离线可看

Community社区discord.gg/4xfD9k8AUF — share your first workflowdiscord.gg/4xfD9k8AUF——把你的第一个流程发出来

Next · Thu 3 Sep下一场 · 9 月 3 日（周四）events.openonion.ai — bring a friendevents.openonion.ai——带个朋友来

Stealth browser free for life · credits loaded · you control the final step.隐身浏览器终身免费 · 额度已充好 · 最后一步由你掌控。

Scan before you leave走之前扫一下

Our WhatsApp group — I answer in here.WhatsApp 群——我在群里回。

chat.whatsapp.com/IOMeCdOZLoWJdRwdN373Xb

35 · SET UP REMOTE35 · 设置 REMOTE DESKTOP → QR → PHONE电脑 → 二维码 → 手机

Pair Your Phone in One Minute

一分钟，把手机和电脑配对

STEP 01Open Set up Remote电脑端点 Set up Remote

STEP 02Scan the QR code用手机扫描二维码

STEP 03Confirm your account确认同一账号与工作区

STEP 04Open Remote on mobile手机打开 Remote

Before the demo: update both apps; keep the host awake, online, and signed in.演示前先更新两端 App，并让电脑保持唤醒、联网、已登录。

37 · THE HARD ONES37 · 最难的那种 SAVE THE PAGE · THEN BUILD先存页面 · 再构建

Build a Skill from a Real Page

最复杂的流程：让浏览器照着真实页面把 Skill 建出来

# Example: build a “linkedin-easy-apply” skill

Some flows are too fiddly to hand-write — a job application
with several steps, a resume upload, a Review screen. For
those, don't guess the buttons. Let the builder read the
real page and build the skill from it.

1. Open a job that has “Easy Apply” in the ConnectOnion browser.
2. SAVE THE PAGE FIRST — the builder works from the real form:
     co browser save_page_context easy-apply
3. The builder reads that saved HTML/CSS and finds the actual
   fields — name, email, phone, resume upload, the Next button —
   instead of selectors that break next week.
4. It fills each step and screenshots to verify before moving on.
5. It stops at Review. You approve before it ever submits.

Save the page, then build, then run. That order is the trick:
a skill built from the real page keeps working; a guessed one
breaks the moment LinkedIn changes a field.

For the hard, multi-step ones.给最难的多步流程。

Too complex to hand-write? Built from the page itself.手写搞不定？照着页面替你建。

38 · A SKILL TO KEEP38 · 送你一个 SKILL YOUR INBOX, FROM CODEX你的收件箱，交给 CODEX

Take This One Home

这个 Skill 带回家

---
name: my-email
description: Read, reply to and send email — from my own Outlook (co outlook), or from my agent's own address (co email). Use when I say check my inbox, reply to someone, or send an email.
---

# My Email

Two mailboxes, two tools:
- co outlook — my own Outlook account. Real people, real replies.
- co email  — my agent's own address. Signups, automations, tests.

## First run only

  co auth microsoft     # a Microsoft login window opens. Sign in once. Done.
  co status             # shows my agent's email address and my credit balance

Then tell me both, in plain words: what my agent's email address is right now,
and how much credit I have.

My agent's address starts as a long hash — something like
0xcd92510bb6@mail.openonion.ai. I can trade credits for a real name:

  co email name <name-I-want>          # is it free? what does it cost?
  co email name <name-I-want> --buy    # claim it — the price comes off my credits

Check the price with the first command and tell me, together with my balance.
Never buy anything without asking me first.

## co outlook — my own account

  co outlook                          # inbox
  co outlook read 3                   # read #3 (numbers come from the last list)
  co outlook search "<name or email>"  # find an old thread
  co outlook reply 3 "text"           # reply INSIDE that thread
  co outlook send to@x.com "Subject" "Body"
  co outlook sent                     # what I have sent
  co outlook scheduled                # queued to go later
  co outlook send a@b.com "S" "B" --at +2h   # schedule it

## co email — my agent's address

  co email inbox
  co email read <id>
  co email send to@x.com "Subject" "Body"

## Rules

- Search for the old thread first and reply inside it. A reply carries the
  whole history for them; a fresh email makes them go find it.
- Show me the draft. I say send. Never send outreach on your own.
- Plain text: a blank line is a new paragraph. Never write HTML.
- Scheduled Outlook mail can only be cancelled in the Outlook client
  ("Cancel Send") — never by command. So show me before you schedule.

Yours to keep.这个归你了。

Save it. One login at home: co auth microsoft.存下来，回家登录一次：co auth microsoft。

Build YourAI SalesEngine

Know the Room First

What You'll Learn Today

Complete a LinkedIn Engage

Why Engage First

This Is Codex

A Few Codex Moves

Skill = Prompt + Tools

A Real Skill: Sydney Weather

Read it. Copy it. Run it.读一读，复制，跑起来。

Create a Project

LinkedIn Detects the Browser

Install the Stealth Browser

One copy. One paste.复制一次，粘贴一次。

Did It Work?

Your First Browser Skill

One skill. One job.一个 Skill，一件事。

Now Run It

Engagement

Your First Like

Your first like.你的第一个赞。

Run the Like

Make It Yours

Like, Then Comment

A comment worth posting.值得发出去的评论。

Build a Skill from a Real Page

For the hard, multi-step ones.给最难的多步流程。

Run the Comment

Write a Good Skill

Which Model to Run It On

Find & Message

Connect with Aaron

Your first connection.你的第一个连接。

Run the Connect

Find the Right People

Who are you looking for?你要找的是谁？

Run the Search

Skill 1 — Write the Message

Teach it your voice.把你的语气教给它。

Run the Writer

Ten Drafts, Then the Best One

Let it compete with itself.让它自己跟自己比。

Skill 2 — Send the Message

The AI drafts. You hit send.AI 起草，你按发送。

This Skill Wrote Itself

Run the Sender

One Command, The Whole Flow

Skills calling skills.Skill 调用 Skill。

Automate Only After It Is Stable

Prepare Overnight夜间准备

Review in Daylight白天确认

The Computer Still Does the Work

Phone sends手机发送

Computer provides电脑提供

Feedback & Networking

Take the Workflow Home

Pair Your Phone in One Minute

Build a Skill from a Real Page

For the hard, multi-step ones.给最难的多步流程。

Take This One Home

Yours to keep.这个归你了。

Build Your
AI Sales
Engine