Your First Win: Ship Something Real Today

Illustrative · chore to shareable link in one sitting

Why this article exists#

Half of new HQ users report being stuck right after setup. The cure is not more reading. It's shipping one small, real thing, start to finish, today. Below are three fully worked first projects. Pick the one that matches you, copy the commands, and follow along. Each one tells you exactly what to type, what HQ will ask, what "done" looks like, and how long it takes.

Two ground rules before you start:

1. One project per session (a session is one sitting of work with the AI). Open a fresh Claude Code session in your HQ folder for whichever project you pick.
2. Pick something real. A toy example teaches you nothing and impresses no one. Pick a task that actually annoys you.

Win #1 (Operator): Turn a messy recurring task into a deployed report#

Best for: anyone with a recurring "assemble information and send it to people" chore, like a weekly status update, a pipeline summary, or a metrics recap.

Time: 30 to 45 minutes.

You'll turn the chore into a small project, have HQ execute it, and end with a polished web page you can send to anyone.

The commands, in order

1. Plan it. Describe the chore to /prd in plain English. The messier the truth, the better:

/prd a weekly team status report: pull together what shipped, what's blocked,
and what's next, formatted as a clean one-page summary I can share every Friday

/prd will ask you a short batch of questions: what "done" looks like, who it's for, where the source information lives, and (if you belong to any companies, isolated workspaces for a business or team) which company this belongs to. Answer in plain language. When it finishes, HQ has written a PRD (a project spec broken into small, verifiable stories) into a project folder, and tells you the project name and its stories (numbered US-001, US-002, …).

2. Execute the first story. Use the project name /prd just gave you:

/execute-task weekly-status-report/US-001

HQ works through the story: it plans, builds, checks its own work against the acceptance criteria from the PRD, and commits the result. You'll see progress as it goes; it only asks you something when a real decision is needed. If the project has more stories, repeat with US-002, and so on, or save them for tomorrow. One story is enough for today's win.

3. Deploy it. When the report artifact exists:

/deploy

HQ packages the result, publishes it, and hands you a link. If the content is sensitive it will offer a gate, for example a password, so only people you give the password to can open it.

What "done" looks like

A URL you can paste into Slack or email, showing a clean, professional page, plus a reusable project in HQ, so next Friday this is a re-run, not a rebuild. That's the whole point: you've converted a recurring chore into infrastructure.

Win #2 (Anyone): Get your ideas out of your head and onto the board#

Best for: everyone, on day one. The fastest possible win. No questions to prepare for, nothing to get wrong.

Time: 10 minutes.

HQ keeps a project board per company (and for you personally). /idea captures a thought onto that board in seconds; /prd promotes the best one into a real project when you're ready.

The commands, in order

1. Capture three real ideas, one command each:

/idea automate the monthly invoice reconciliation spreadsheet

/idea a one-page FAQ site for the questions customers ask over and over

/idea clean up and standardize our proposal template

For each, /idea may ask one short batched question, typically which company the idea belongs to, and whether it's tied to a specific project or repo. Answer or skip; the idea lands on the board either way.

2. Promote the best one:

/prd clean up and standardize our proposal template

Same /prd flow as Win #1: a short interview, then a PRD with numbered stories, ready to execute now or later with /execute-task.

What "done" looks like

Three ideas safely on your board instead of in your head, and one of them turned into an execution-ready project. From now on, when inspiration hits mid-meeting, you have a two-second home for it. (Wondering when to use /idea vs /prd vs the heavyweight /deep-plan? Rule of thumb: /idea = capture a one-liner, /prd = standard project planning and right for ~90% of work, /deep-plan = big strategic efforts only. More in idea-to-shipped.md.)

Win #3 (Builder): Point HQ at a codebase you already have#

Best for: technical users with an existing repo. This is how HQ goes from "generic assistant" to "knows my codebase."

Time: 20 to 30 minutes, and most of it is hands-off while HQ explores.

/discover pulls a repo into HQ, fans out parallel exploration across it, and writes structured, searchable knowledge about the codebase (architecture, conventions, how to build and test) under the owning company.

The commands, in order

1. Start a fresh session (discovery is context-hungry; give it a clean slate), then:

/discover your-org/your-repo

It accepts GitHub org/name shorthand, a full git URL, or an absolute path to a checkout you already have locally. Useful flags: --company <slug> to file the knowledge under a specific company, --private to force placement under repos/private/.

2. Answer the gates. /discover will confirm where to put the repo (repos/public/ or repos/private/), which company owns it, and (near the end) whether to adopt any proposed policies (standing rules HQ follows, e.g. "always run the linter before committing"). It only proposes policies for conventions that are crystal-clear in the codebase, and nothing is adopted without your yes.

3. Prove it worked. In a fresh session, ask something only your codebase can answer:

/search how does authentication work in your-repo

What "done" looks like

The repo lives under repos/, and HQ has written knowledge files about it that every future session can search. Ask an architecture question and get a specific, correct answer with file references. That's the payoff. From here, /prd a real change against the repo and execute it story by story.

Now make it a habit: the daily loop#

Whichever win you shipped, close it out properly. This is the habit that separates people who compound with HQ from people who plateau:

1. End today's session with /handoff. It saves a handoff (a structured snapshot of where you left off: state, next steps, open threads).
2. Tomorrow, open a fresh session with /startwork. It reads your last handoff and your board, and offers you the next most useful thing to do.
3. Repeat. Fresh session per task, /handoff at the end, /startwork at the start.

That loop is the engine of HQ. Read daily-loop.md next. It's short, and it's the last piece of "how do I actually live with this thing."

Common questions#

If you get stuck#

The most common snag is running commands from the wrong place. Make sure Claude Code is open in your HQ folder. If a command errors mid-flow, start a fresh session and try again; HQ picks up saved state. Decision tree for everything else: im-stuck.md, or report it with /hq-bug.