Skip to content
The Liberation Engine

Things to make / Recipe 05 of 14 · Play

A story you can walk through

Open a locked museum after closing and choose which door to trust.

What the agent made

Ran it for real with Claude Code 2.1.292 and OMP 18.8.5 on 2026-10-08, in an empty session with no saved memory. Your result will differ.

Open what Claude Code made, full screen

The same prompt in OMP made its own version: open what OMP made. Two runs, two different results: that is normal.

Make your own

  1. Get the files

    Make an empty folder called cli-practice in Downloads. No downloads: start in an empty folder.

  2. Open your agent in that folder

    Open a terminal in cli-practice and type claude or omp --approval-mode always-ask. First time? Set up in about five minutes, with steps for Mac, Windows, and Linux.

  3. Paste this prompt

    Create seven offline HTML pages for a branching story about an unopened museum after closing: index.html, gallery.html, archive.html, basement.html, ending-a.html, ending-b.html, ending-c.html. Start links to gallery and archive. Each non-ending page offers two consequential choices; all three distinct endings must be reachable. Every ending links back to index.html. Use relative links, self-contained pages, and readable text. Check every link resolves locally. Work only inside this practice folder; never edit inputs, install packages, call APIs, send messages, or use external assets. Use no em dashes or en dashes.

    The prompt asks the agent to stay in this folder; it cannot enforce that. Read each file change and command before you approve it.

  4. Check it yourself

    Open cli-practice/index.html in your web browser (double-click it), then:

    Human check: Follow two paths to different endings, return to the start, and visit every link to confirm all destinations exist.

    When we checked the recording above: Opened each of seven HTML pages in browser and clicked all 11 local links per recording; every destination loaded with a heading. Played routes to each of three distinct endings and clicked return-to-start links. No external requests.

Why an agent, not a chat window

An agent creates a navigable set of real pages and inspects links across files, rather than offering disconnected story text in chat.