Solution entry · Agents and AI

Prompt Escape

A multiplayer prompt-engineering escape room that turns clearer instructions, constraints, iteration, and creativity into a playable team challenge.

Published 2026-09-14 Updated 2026-09-14 Repository created 2026-02-12 Last public push 2026-03-06 Public hackathon prototype
Original state to outcome

What changed in the design.

Customer-neutral framing keeps the pattern reusable without claiming private results.

01 / Original state

Before the solution

Prompting concepts can stay abstract when learners only read examples instead of working through a shared problem with constraints, feedback, and room to iterate.

02 / Delivered

What was built

The public project delivers a TypeScript and Express escape-room game with solo, cooperative, and versus modes, deterministic prompt scoring, hints, and themed challenges.

03 / Outcome

Value enabled

The experience is designed to make prompt quality discussable and memorable through practice, comparison, and team reflection.

Architecture

How the public pattern fits together.

A team challenge leads to a prompt attempt, game feedback, and team reflection before another iteration.

Prompt Escape architecture A team challenge leads to a prompt attempt, game feedback, and team reflection before another iteration. 01 Team challenge 02 Prompt attempt 03 Game response 04 Team reflection
01 / Team challenge

Players enter a shared scenario with a goal, clues, and constraints.

02 / Prompt attempt

The team writes and refines instructions for the challenge.

03 / Game response

The application returns challenge feedback and progression state.

04 / Team reflection

Players compare approaches and decide how to improve the next attempt.

Capabilities

What the project demonstrates.

  • Packages prompt-engineering concepts as a multiplayer game.
  • Uses clues and constraints to reward clearer instructions and iteration.
  • Supports team play and discussion rather than an isolated tutorial.
  • Uses keyword, synonym, and heuristic scoring rather than an external AI evaluation API.
Human control and guardrails
  • The game is a learning experience, not a benchmark of a person or model.
  • Facilitators should avoid entering sensitive, personal, or proprietary information into prompts.
  • AI responses are part of the challenge and should be treated as fallible output.
Limitations
  • The WorkIQ hint service currently returns mock responses rather than making live MCP calls.
  • The experience does not measure durable skill improvement or workplace outcomes.
  • The repository-listed hosted site returned HTTP 403 during verification, so current public hosting is not established.
Public provenance

Follow the evidence.

These public sources support the entry. No private customer material or internal evidence was used.