Step 3 of 6

Populate

One room is not much of a labyrinth. Let's fill it out.

Ask for more rooms

You have one room and a template the agent knows how to follow. Now ask it to build the rest of the maze:

Tell your agent

"Create four more rooms for the labyrinth. I need:

• Twisting Corridors — narrow stone passages, claw marks on the walls, bones on the floor
• Underground Pool — still dark water, dripping stalactites, echoes
• The Minotaur's Lair — the center of the maze, a beast waiting, golden treasure scattered
• Ariadne's Chamber — a window overlooking the sea, a spool of thread, hope

Connect them: Entrance Hall leads to Corridors, Corridors leads to Pool and Lair, Pool leads back to Corridors, and Ariadne's Chamber connects to the Entrance Hall."

The agent creates four pages, all following your template. Each one has ## Description, ## Exits, ## Notable Features — without you mentioning the template at all.

See the whole world

Tell your agent

"List all the txtscape pages."

The agent calls list_pages and shows:

📄 rooms/ariadnes-chamber.txt  —  # Ariadne's Chamber
📄 rooms/entrance-hall.txt  —  # The Entrance Hall
📄 rooms/minotaurs-lair.txt  —  # The Minotaur's Lair
📄 rooms/twisting-corridors.txt  —  # Twisting Corridors
📄 rooms/underground-pool.txt  —  # Underground Pool

Five rooms, all in the rooms/ folder, all following the same template. The labyrinth has a shape.

💡

Paths you can use directly. Every path in the listing (like rooms/ariadnes-chamber.txt) is a valid argument to get_page, str_replace_page, or delete_page. The agent copies a path from the listing and uses it immediately — no guessing, no path manipulation.

🔬

Try this. Start a new chat session (close and reopen the AI chat). Ask the agent "What rooms exist in this labyrinth?" — it can call list_pages and find everything, without you re-explaining anything. That's persistent memory at work.

Commit your progress

git add .txtscape
git commit -m "populate five labyrinth rooms"

← Define Deepen →