Rapid Build Hackathon
Viksit UP · AI Challenge
Room 02 · Prompt Library

From pain point
to a runnable pilot prompt.

Six prompts, designed to be run in order. Each one refines the thinking from the last — from a grounding research brief, through diagnosis and ideation, to a brief and finally a platform-ready build prompt. Use them with any frontier LLM.

How the flow works

Paste the Learn prompt to ground the model in the topic, then move through Widen and Diagnose to understand the problem. Ideate converges on a pilot concept. Brief turns it into a one-pager. Build writes the product requirements prompt you ship to a build platform (Cursor, v0, Bolt, Replit Agent, etc.).

Tip: replace bracketed tokens like [selected challenge] with the challenge you picked from Room 01.

Prompt sequence

01Learn↓
02Widen↓
03Diagnose↓
04Ideate↓
05Brief↓
06Build→
STEP 01 / GROUND

Learn

Feed the model background context before anything else.

context-seed no-action attach-report
Here is a deep research report for the topic we are going to discuss today. No action required. Use this report as additional context for your responses apart from web search and other resources.
PASTE FIRST · ATTACH REPORT
STEP 02 / DISCOVER

Widen

Map personas, pains, workarounds and success metrics.

research-aide 5 insights · 3 risks
Act as a research aide for [selected challenge]. List key personas, top pains, current workarounds, and success metrics. Return 5 insights & 3 risks tailored to this challenge context.
REPLACE [selected challenge]
STEP 03 / ROOT-CAUSE

Diagnose

Pick the top pain and run a structured Five Whys.

five-whys test-plan privacy
Let's pick the [top-pain-point] for this challenge. For this pain, run a Five Whys. Propose 3 root-cause hypotheses and the disproof evidence for each. Specify the minimum data cut & owners to pull. Output a root-cause map, test plan, and privacy constraints.
REPLACE [top-pain-point]
STEP 04 / CONVERGE

Ideate

Cluster AI-driven ideas. Score. Recommend a pilot.

process analytics/ML AI & automation impact × feasibility × confidence × TTV
Generate and cluster possible AI-driven ideas into 3 options: 1. Process (policy, ways of working) 2. Analytics / ML (forecast, optimise, recommend) 3. AI & Automation (Computer Vision, Retrieval-Augmented Generation / Agentic AI, etc.) Score each on Impact × Feasibility × Confidence × Time-to-Value. Recommend one pilot with the smallest integration surface and the clearest value proof.
CONVERGE TO ONE PILOT
STEP 05 / CONDENSE

Brief

Turn the chosen pilot into a one-page pilot brief.

one-pager metrics & baselines guardrails
For the recommended pilot, create a one-page pilot brief including: target user(s), problem statement, success metrics & baselines, target uplift, key flow (5–7 steps), screens / components, sample UI copy, representative sample data, integration points, and relevant guardrails (GDPR / PCI, domain-specific regulation boundaries, bias tests, fallback behaviour).
ONE PAGE · CONCRETE
STEP 06 / SHIP

Build

Generate the platform-ready product requirements prompt.

PRD-as-prompt screens + components no PII
You are a product design expert. Using only the brief above, write a single [platform] product requirements prompt that includes: product name + one-liner description (actions, process, capabilities), who it's for, screens + key components, brand colours, main user flow, sample data, concise headlines / CTAs, UI instructions, a success-metric card, and constraints (no PII). Return the [platform] prompt only.
REPLACE [platform] · e.g. v0, Cursor, Bolt