Lesson 7 of 7 / HQ fundamentals

Create, Train, and Share a Reusable HQ Worker

Create a narrow company worker, inspect its durable files, enrich it with reusable design knowledge, and share it as living team infrastructure.

Begin lesson

Complete in: 10–15 minutes
Written path: 9 steps

Outcome

You will finish with a company-scoped worker whose skill loads a saved aesthetic pack and can improve as teammates add better knowledge and examples.

Recorded walkthrough

Watch the complete path.

Follow the recording once, then use the written steps below as the durable checklist. English captions are available from the player controls.

HQ / Tutorial recording

This walkthrough uses demo data and shows the complete recorded workflow. Use the written steps below as your durable checklist while you follow along.

Before You Begin

HQ is installed and you can start a company-scoped session.
You have permission to create and edit workers for the company.
A public reference website is available for the aesthetic-pack exercise.
The teammate exists in HQ if you plan to complete the optional sharing step.

Step 1
Choose a narrow specialist
Define one repeatable job rather than a general-purpose agent. The recording uses a landing-page generator to show how a worker accumulates role, skills, knowledge, and examples.
Expected result
You can explain the worker's repeatable responsibility and how it differs from one static skill.
Step 2
Create the worker in company scope
Enter the target company, then invoke the worker scaffolder with a descriptive slug.
CommandRun exactly as shown
```
> /startwork awesomeco
> /newworker landing-page-generator
```
Expected result
HQ opens the company worker scaffold and asks configuration questions.
Field note
awesomeco is the recording's example handle. Use the company where the worker should actually live.
Step 3
Configure scope, output, and delivery
Keep the worker company-only, choose deployable-page code as the output, and make it available on demand.
SelectionChoose in the interface
```
— AwesomeCo only
— Deployable page (code)
— On-demand
```
Expected result
HQ creates and registers the company-only Landing Page Generator.
Field note
AwesomeCo is the demonstration company; select your own company scope in a real setup.
Step 4
Inspect the generated worker files
Open the worker directory and review its definition, readme, and generated landing-page skill before adding more knowledge.
NavigationOpen this destination
```
→ worker.yaml
→ README.md
→ generate-landing-page.md
```
Expected result
You can see the worker metadata, run configuration, and referenced skill.
Field note
The files live under companies/[company handle]/workers/landing-page-generator/; replace the bracketed handle.
Step 5
Request a reference aesthetic pack
Ask HQ to inspect a public reference site and save specific design guidance for this worker.
PromptSend in your AI session
```
› I want this worker to have more specific design instructions. Please reference hqforwork.com and build a design and aesthetic pack this worker can use.
```
Expected result
HQ inspects the reference site and asks how closely future pages should follow it.
Field note
hqforwork.com is the public example from the recording; choose a site you are allowed to reference.
Step 6
Choose the fidelity
After the site inspection finishes, select the degree of visual fidelity appropriate for the worker.
SelectionChoose in the interface
```
— Select Match it closely
```
Expected result
HQ writes the aesthetic pack and integrates it with the worker instead of returning a one-off analysis.
Step 7
Verify durable worker context
Confirm the pack is saved in company knowledge and that both the worker definition and skill reference it.
NavigationOpen this destination
```
→ knowledge/design/hq-aesthetic-pack.md
→ worker.yaml
→ generate-landing-page.md
```
Expected result
Every future worker run can load the saved aesthetic pack as shared base context.
Step 8
Share, notify, and hand off
Ask HQ to share the worker with the intended teammate, send an HQ DM, and hand off the session without placing private identity data in reusable copy.
PromptSend in your AI session
```
› Share this worker with [teammate], send an HQ DM, and hand off.
```
Expected result
The authorized teammate can use the same worker, skill, and reference pack, and the session is saved.
Field note
[teammate] is a placeholder. Use HQ's private identity flow rather than publishing contact details.
Step 9
Improve the worker over time
Treat the worker as living team infrastructure: refine its skills, pack, and examples whenever the team learns a better way to do the job.
Expected result
Future runs benefit from accumulated team knowledge instead of a frozen one-session prompt.

Observable finish line

You are done when this is true.

The company worker exists with its definition and skill, the saved aesthetic pack is referenced by future runs, and an authorized teammate can use and improve the same reusable worker.