One prompt. One review-ready result.

Run this Loop It goal in Codex as one end-to-end task. Follow the instructions directly as a normal agent task.

Loop goal: First Value Sprint
Category: Product & UX
Goal: Walk through the first-time user journey, identify the single biggest barrier to reaching value, make the smallest credible improvement, and prove the journey is clearer on desktop and mobile.

Workspace boundary:
Work only with the source material and workspace available in this task. Do not assume this is a software repository or that code changes are required. Keep the scope to one reviewable artifact or improvement.

Required inputs:
- Product surface: The local app, prototype, screenshots, or flow to review. (required)
- First value event: The user action that proves the product became useful. (required)

Expected artifact:
A focused product change or review-ready recommendation with before-and-after journey evidence.

Expected result: Review-ready deliverable
Evidence to capture: Rubric evidence
User proof requirement: Return Journey notes, Before-and-after evidence, Remaining risks that demonstrates: The first value event is stated in observable terms. The highest-impact friction point is supported by inspected evidence. The revised journey completes at desktop and mobile widths without introducing a new dead end.
Proof mode: hybrid
Iteration cap: 3

Workflow: UNDERSTAND -> CREATE -> CRITIQUE -> REFINE -> PROVE.

1. UNDERSTAND — inspect the supplied material, identify the audience and decision, and list missing inputs without inventing them.
2. CREATE — produce the smallest complete version of the expected artifact using only supported claims and available evidence.
3. CRITIQUE — review the artifact against every rubric criterion and name the most consequential gap.
4. REFINE — make one focused pass that clearly improves the rubric result without expanding scope.
5. PROVE — return the artifact, criterion-by-criterion evidence, assumptions, blockers, remaining risks, and the next safe action.

Rubric criteria:
- The first value event is stated in observable terms.
- The highest-impact friction point is supported by inspected evidence.
- The revised journey completes at desktop and mobile widths without introducing a new dead end.

Approval required for:
- production writes
- external messages
- customer data access
- deploys
- irreversible changes

Do not ask me to run, copy, or paste terminal commands. Perform any necessary local inspection inside the agent workflow and keep the user-facing exchange in natural language.
Do not publish, contact customers, spend money, change accounts, write to production, deploy, or take an irreversible action without explicit approval.
Treat the deliverable as review-ready when every rubric criterion has evidence. Review-ready means the requested work is complete for review; it does not mean approved, published, or deployed.

Stop when the rubric evidence is complete, 3 iterations are reached, required input or access is missing, the same gap repeats twice, approval is needed, or an unsafe action would be required.

Return the review-ready deliverable with criterion-by-criterion evidence. If completion is blocked, return the blocker, the missing input or approval, and the safest next action.

Runs inside your local workspace.