> ## Documentation Index
> Fetch the complete documentation index at: https://docs.aiml-code-it.automat-it.io/llms.txt
> Use this file to discover all available pages before exploring further.

# End-to-end walkthrough

> Build a project from an Asana ticket, then score the run.

This is a full run from ticket to pull request, with a scorecard. Use it to check that the whole chain works after a release.

> [!NOTE]
>
> Items marked **Unverified** were not confirmed on a live deployment.

## The flow

**1. Write the ticket**

Put context, scope, an acceptance checklist, constraints and subtasks in the Asana task. Describe the work, not the tooling: do not name skills. Nobody knows the skill names in advance, so the agent reads each skill's description and decides for itself which ones fit. Afterwards, the **Usage** page shows which skills it chose. The agent reads the task through read-only Asana tools and cannot change it.

**2. Create the agent from the ticket**

Open **Agents**, then **New agent**. Pick the project and the task. The agent starts as `waiting_for_runner` until your laptop picks it up.

**3. Choose a work profile**

If the project defines presets, pick one in the dialog. It sets the plan-approval gate, the reviewer gate, test depth and expected skills. See [Work profiles](https://docs.aiml-code-it.automat-it.io/guides/work-profiles.md).

**4. Plan and answer questions**

When the agent asks something, it parks and the chat shows a question panel. Pick an option or type an answer. See [Chat features](https://docs.aiml-code-it.automat-it.io/guides/chat.md).

**5. Approve**

Approve the plan if the profile asks for it. Do not approve any `code-it-apply` request for a ticket that forbids deploys.

**6. Steer**

Change model and effort in the composer. The choice applies to that chat only. Corrections you type are recorded as interactions.

**7. Tests and draft PR**

The agent runs the project checks and the PR hook opens a pull request on the ticket branch. Confirm it is a draft.

**8. Review and keep corrections**

If you corrected the agent and it then stops at an approval, a card asks whether to keep the correction. Yes writes a lesson. See [Lessons](https://docs.aiml-code-it.automat-it.io/guides/lessons.md).

**9. Read the metrics**

Check Usage, Interactions, Cost efficiency, Delivery velocity and DORA. See [Cost and metrics](https://docs.aiml-code-it.automat-it.io/guides/cost-and-metrics.md).

## Before you start on the laptop

* Run `code-it-runner status`. It should be online.
* Run `code-it-runner doctor`. It should report no problems.
* Check `kiro-cli whoami` and `gh auth status`.
* Confirm the project's PR hook opens drafts.
* Open **Agents**, **New agent**, pick the ticket and start.

## Scorecard

Score the run with the [scorecard](https://docs.aiml-code-it.automat-it.io/guides/scorecard.md). It also explains how to read the Usage page and its `.code-it/usage-events.jsonl` sink.

## Unverified

* Live numbers on the metrics pages on the current deployment.
* Whether the agent reliably runs and reports the project check. Read the timeline each run.
* Whether a ticket lands in the section Code-it treats as ready. That depends on the project's Asana launch settings.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.