Compete in a real-time strategy game
A game environment built for frontier LLMs to play against each other.
Done with a custom-built or unnamed agent
- What they did
- LLM Skirmish is a set of benchmarks where LLMs write code that plays strategy games. In the Screeps version, models write JavaScript bots for a Screeps-style real-time strategy game. The bots then fight in tournaments and on a community ladder. A StarCraft: Brood War version, StarSkirmish, has LLMs write bots for that game.
- How it went
- The page gives no scores or numbers. It links to bench results and a hillclimb livestream for StarSkirmish, and the Screeps page covers the tournaments and ladder.
- Worth knowing
- The home page is thin on mechanics, such as setup, cost and how much a human steps in. Check the Screeps and StarSkirmish pages before copying it.
Try it yourself with Claude Code
Help me set up an LLM-playable real-time strategy game environment, such as [game or repo name]. Write an agent that reads the game state each turn and issues commands to [goal, e.g. win a match against a scripted bot]. Finished means one full match has been played and you've given me a short log of what went right and wrong.
1 more account of this
Discussion on HN · Feb 25, 2026
Five of these in your inbox every morning
The best things people got an AI agent to do, each with the prompt to try it.
More like this
Play RollerCoaster Tycoon
Ramp Labs put Claude Code in RollerCoaster Tycoon.
- Custom agent
Test a browser agent on web games
Researchers had ChatGPT Atlas's browser agent play T-Rex Runner, Sudoku, Flappy Bird and Stein.world; it did well on logic puzzles like Sudoku, finishing faster than humans.
Build Godot games with paired design and coding agents
Cowork handles design docs, task queues and playtests while Claude Code writes GDScript and runs Godot and git.
Build a racing game from a single prompt
Codex made a racing game with multiple racers, eight maps and items from one prompt, acting as designer, developer and QA tester by playing it.
Navigate Goat Simulator 3 with a game-playing agent
Gemini-powered SIMA 2 navigates Goat Simulator 3, and a human player can direct it to improve its performance.
Build a minigame collection in a day
A data professional drafted a prompt with ChatGPT and used Cursor agent mode to build a Python tkinter game with four minigames.