---
title: Quick Start
description: Commands to run after installation, ordered from planning to merge
seo:
  image: /ogp.png
---
Once installation is complete, you can walk through the entire workflow from planning to merge by running the commands on this page in order.
Every command can be pasted directly into the Claude Code prompt and executed there (not in your shell — in Claude Code).

For a detailed explanation of what each skill does, see [Skill Details](/skills/product-pipeline).
We recommend running through this first to get a feel for the flow, then reading the details afterward.

## 1. Plan your product

```
/product:start --profile=mvp
```

`--profile=mvp` is the shortest path: it takes you through defining your <Tooltip tip="A statement of what you want to achieve with this product. The starting point of the pipeline." headline="Vision" cta="See in glossary" href="/glossary">Vision</Tooltip> and <Tooltip tip="What is in scope and what is out of scope for this iteration. Determined using MoSCoW or RICE scoring." headline="Scope" cta="See in glossary" href="/glossary">Scope</Tooltip> all the way to a Go/No-Go decision.
For your first run, this profile is the recommended way to get a feel for the overall flow.

| Profile | Phases | When to use |
|---|---|---|
| `mvp` | 3 | Just need a Go/No-Go decision quickly (start here) |
| `core-only` | 6 | Want to solidify the core business case |
| `ux-to-spec` | 9 | Vision/Scope already defined; want to drive to a spec |
| `full` | 23 | Want to go end-to-end from Vision to architecture (default) |

Output is written under `reports/`. See [Output Directory Structure](/getting-started/output-structure) for details on the layout.

## 2. Create the backlog

```
/architect:export-backlog --dry-run
```

With `--dry-run`, no Issues are created — only the plan file `reports/backlog/backlog-plan.md` is written.
Use this to verify that the Epic / Sub-Epic / Issue breakdown matches your intent before committing.

Once you are satisfied, run it for real by dropping `--dry-run`:

```
/architect:export-backlog
```

This creates the actual Issues in GitLab/GitHub.

## 3. Implement

```
/architect:implement-backlog
```

If no item is specified, the skill picks up items with `status::doing` and asks for confirmation before starting.
To target a specific Issue, pass its ID: `/architect:implement-backlog I1.1.1`.

This skill commits the code but **does not create a PR/MR**. That happens in the next step.

## 4. Review and open a PR/MR

```
/architect:review-issue
```

If blockers are found, the skill automatically cycles through "fix → re-review" (up to 3 times by default).
Once the blocker count reaches zero, the skill creates the PR/MR.

## 5. Merge

```
/architect:merge-issue
```

The merge will not proceed unless every preflight check passes.
You will always be prompted for confirmation immediately before execution.

## Running everything at once

Instead of running steps 3–5 one at a time, you can run them together at the Epic level:

```
/architect:deliver-backlog --epic=E1
```

This runs `implement` → `review` → `merge` for each Issue in sequence. It is not fully autonomous, however.
The workflow pauses at any point that requires a human decision — PR/MR approval, merging, and blocker resolution.
Progress is recorded in `backlog-manifest.json`, so if you interrupt the run you can resume it with the same command.

:::warning[Always run hard-to-reverse operations yourself]
`export-backlog` (Issue creation) and `merge-issue` (merging) modify external trackers and repositories and are
hard to reverse. The major skills that include these operations have `disable-model-invocation` set, meaning
**they will not execute on Claude's judgment alone**. They only run when you enter the slash command yourself.
The reasoning is explained in [Installation](/getting-started/installation).
:::

:::note[If you get stuck]
Issues encountered in practice — such as labels multiplying unexpectedly or Epic features not working —
are documented in [Troubleshooting](/troubleshooting).
:::