Learn Autohand Code at your own pace

This self-paced learning path teaches Autohand Code through a small task-list project with no application dependencies. You will explore source code, fix an intentional bug, add a tested feature, prepare a handoff, and automate validation. Each lab has a checkpoint; proceed when you can demonstrate the result.

Prerequisites and estimated time

Use Node.js 20 or newer, npm, Git, and a working Autohand Code installation with a configured provider. Start with your first session if needed. Agent use depends on your account or provider and can incur usage. The local fixture and its tests require no package download.

Allow approximately 90–120 minutes for the complete path; these are learning estimates, not measured completion times. You can stop at any checkpoint and resume in the same practice folder. No deployment, external service integration, or production data is needed.

Choose your next lab

Lab	Estimated time	What you will demonstrate
Set up and explore	15–20 minutes	Identify an intentional failure and trace it to a function
Fix a bug	20–25 minutes	A focused repair with meaningful regression tests
Add a feature	20–25 minutes	An explicit contract and tests for boundary cases
Review and hand off	20–25 minutes	A reviewed diff and a resumable evidence record
Automate a check	15–25 minutes	A validation command that fails when a requirement breaks

How to use the labs

Shell blocks run in your terminal. Prompt blocks go into an interactive Autohand session. Read each prompt before submitting it and adapt only the parts the lab asks you to change. The agent's wording and implementation may vary; compare behavior and test results, not a screenshot or exact answer.

The starter intentionally has one failing test. Do not ask the agent to fix it until you have recorded the failure in the exploration lab. Later, add tests that would catch the original bug rather than only confirming the new implementation.

Pause, resume, and reset

Save a short note with the lab completed, current test result, and next step. Before resuming, run git status --short and npm test to re-establish the current state. To start over, download the fixture into a new empty directory; keep your existing work for comparison.

Final assessment

Without copying a lab prompt, add a renameTask(tasks, id, title) function with an explicit policy for blank titles and unknown IDs. Plan its behavior, write a failing test, implement it, and review the diff. Finish when you can explain why the tests establish the chosen contract and identify any behavior you did not check.

Next learning paths

Apply the workflow to a real repository using codebase exploration, monorepo workflows, and context handoffs. For programmable agents, continue with the Code Agent SDK quickstart.