Handbook

Wizard 301 — Advanced usage

Deeper use of the Wizard: artifact bundles, Refine and LLM-assisted steps, review and recheck, and Cursor Launch Pack packaging. This hub links to focused pages; read the worked example session below once so the advanced…

What it is

Deeper use of the Wizard: artifact bundles, Refine and LLM-assisted steps, review and recheck, and Cursor Launch Pack packaging. This hub links to focused pages; read the worked example session below once so the advanced topics read as one workflow, not isolated toggles.

When to use it

After Wizard 201, when you are tuning outputs or exports.

Prerequisites

  • Stable server-enabled sessions when possible (Wizard 201).

Topics

Topic Page
Artifact bundles Artifact bundles
Refine Refine
Review and recheck Review and recheck
Cursor Launch Pack Cursor Launch Pack

Worked example session (generic)

Scenario: A squad runs a Wizard session for an internal API reliability initiative — not greenfield, but enough discovery that Mission and Understanding still matter.

1. Bundle choice (Artifact bundles) — At Target & output pack, the team picks Engineering emphasis (labels may say “engineering” or similar in your build). Why: they need tasks, risks, and guardrails for the next increment more than a marketing-style roadmap slice.

2. Refine (Refine) — In Understanding, they run Refine on a short paragraph about current outages. The first pass smooths wording but drops an explicit SLO mention. The product lead edits the notes to put the SLO back, adds a non-goal (“no new datastore”), and Refines once more. The second pass keeps both.

3. Review and Recheck (Review and recheck) — Review & generate surfaces an artifact set that lists two owners for the same risk class. Recheck flags an inconsistency between the run plan and the risk table. The team fixes the session (single owner per risk class, dates aligned), regenerates, and Recheck passes.

4. Cursor Launch Pack (Cursor Launch Pack) — At Experimental build / handoff, they export a Launch Pack (wording varies). At a high level it includes: a session summary slice, near-term tasks with owners, and constraints copied from notes — enough for an editor session without pasting private chat. They treat it as context for humans, not auto-execution.

Outcome: One coherent story from emphasis → Refine → gate → export; child pages below unpack each lever.

How to verify success

  • Exports match your policy; recheck passes or you knowingly accept gaps.

What to do next