Demo Script

Demo Script

Use this script for a short zero-to-first-success demo.

The first AIWG demo should show one path working, not a tour of every framework. The success metric is small on purpose: one disposable project, one wizard preview, one deploy, one status probe, and one recommended next action.

We chose a disposable project because demo failures should be cheap. While a 300s walkthrough is not a full tutorial, it is enough to catch the common problem: the user is in the wrong folder, the provider was not chosen, or the probe is not ready.

Validation baseline: test this script against AIWG version 2026.5. While the demo is intentionally small, maintainers should record any issue that failed before the first status probe.

Setup

Use a disposable project with no private data.

mkdir aiwg-demo
cd aiwg-demo
git init

Script

1. Show the goal.

I want AIWG to help me start a small project without learning the whole system first.

2. Preview the wizard.

aiwg wizard --dry-run --goal "help me start a project"

Point out that no files were written.

3. Run the guided path.

aiwg wizard

In automation, use:

aiwg wizard --non-interactive --profile beginner --provider codex

4. Verify engagement.

aiwg status --probe --json

Expected result: `engaged` is `true` after a successful deploy, or the probe gives the next command to run.

5. Ask for one first action.

Recommend one AIWG path for this demo project, one reason, and one fallback. Use aiwg discover and aiwg show before recommending.

6. Stop after the first useful output.

Good demo ending:

AIWG is installed in this project, the provider files are present, the status probe is ready, and the next action is a single intake or discovery path.

Recovery Beats

If the demo starts in the wrong folder, use:

pwd
ls -a
aiwg status --probe --json

Then move to the project root and rerun the wizard preview.

If provider behavior is uncertain, say:

The local deployment is verified by the probe. Provider-session behavior is still tracked separately in the field-validation matrix.

This wording avoids a common problem: deployment evidence and provider-session behavior are different evidence types.