Game Dev with AI Agents: method, prompts, skills and check loops
by PaperChaseWebb
A Markdown playbook for building games with Claude Code, Codex or Claude: a seven-stage method, a staged brief, 16 prompts, five agent skills, self-check loops, engine notes and 30 game concepts.
- Price
- $39
- Difficulty
- Intermediate
- Setup
- 20 min
- Format
- .md file
game-dev-with-ai-agents.md · 37 KB · instant download after checkout
Want this built for you instead? We can build it for you.
What you end up with
Your agent proves each milestone works with play logs, frame sheets and benchmark numbers before it tells you it is done.
What you get
- 11 sections in one .md file, about 25 pages
- Seven-stage method and a one-page staged brief template
- 16 stage prompts from brief to release gate
- Five SKILL.md files: playtester, frame critic, perf runner, asset smith, release gate
- Self-check loops: agent plays the game, frame review, benchmark harness, headless simulation
- Asset-as-code recipes and an asset license log
- 48-hour jam checklist with cut lines
- Engine notes for Godot 4, Unity 6, Three.js and Unreal 5
- Concept library of 30 games with risks and first prompts
- Linked credits for all 62 sources we studied
Who it is for
Indie developers, jam entrants and first-time game makers who use an AI coding agent and want it to test its own work instead of guessing.
Works with
Solid chips: the agent is named in this listing's own tools list. Dashed chip: Plain Markdown, so it also pastes into other agents such as Grok Bot, Hermes, OpenClaw or Muse. We list those as compatible by format only, not as tested.
Tools needed
Cost to run: Free. Uses the agent plan you already pay for.
Time to result: Brief and first playable in one evening
Sample prompts, free to use
Play the game using the same inputs a player uses (key presses or taps sent to the game window), not by setting positions in code. Complete one full round. Write PLAYLOG.md with what you pressed, what happened and every bug. Fix the bugs, replay, and only then report done.
Scripted tests can pass while real controls are broken.
Using BENCH.md as the baseline, try one optimization at a time. Rerun the benchmark after each. Keep only changes that improve worst frame time without breaking PLAYLOG checks. Stop after [n] attempts or when under budget.
Give the agent a number it can move, and a stop rule.
Free preview of the file
# Game Dev with AI Agents A PaperChaseWebb pack, version 1.0. Written for Claude Code, Codex and Claude. Plain Markdown, so it also reads fine in any other agent, but those three are the ones every prompt and skill was tested against in wording and structure. This pack does not build a game for you. It gives you the method, briefs, prompts, skills and check loops that the builders in our October 2026 research used to get from idea to a playable, testable game with an agent doing most of the typing. We studied 62 public build reports (linked in `10-sources.md`) and wrote everything here in our own words. ## Files | File | What it is for | |---|---| | 01-method.md | The seven-stage loop and the rules that make it hold up | | 02-staged-brief-template.md | A fill-in game brief split into milestones an agent can finish one at a time | | 03-prompts.md | 16 stage prompts from first build to portal upload | | 04-agent-skills.md | Five drop-in skill files: playtester, frame critic, perf runner, asset smith, release gate | | 05-self-check-loops.md | Agent plays the game, frame review, benchmark harness, headless simulation, builder and critic split | | 06-asset-as-code.md | Art, sound and models generated by code, plus a license checklist for anything that is not | | 07-jam-checklist.md | A 48-hour jam plan with cut lines | | 08-engine-notes.md | Godot 4, Unity 6, Three.js and Unreal 5 notes, with the traps agents fall into | | 09-concept-library.md | 30 game concepts with the core loop, the agent risk and the first prompt | | 10-sources.md | Every source we read, linked and credited |
Worth knowing
- The pack is a method and prompt library. It does not include a game, engine or assets.
- Engine APIs change. Confirm versions against official docs.
- Never let an agent upload, submit or spend without asking you.
Questions
Which agents is it written for?
Claude Code, Codex and Claude. The skills use their file layouts. Other agents can read the Markdown, but we have not written for them.
Is there a free sample?
Yes. The playtester rule and skill are free on their own listing.
Is any of it copied?
No. We read 62 public build reports and wrote everything in our own words. Sources are linked and credited inside.
Details
Version 1.0 · Updated Oct 25, 2026 · Markdown version of this page
License: Personal and commercial use. Your games are yours. Do not resell or republish this pack. Download terms
Related
Make your agent play the game before it says done (free sample)
One rule, one playtester skill and a play log template that make Claude Code or Codex test your game with real controls before reporting done.
Build a playable browser game prototype in one evening with an AI agent
A step by step pack for going from a one-paragraph idea to a playable HTML5 game in a single file, then posting it on itch.io for friends to try.
Launch an AI-built browser game: mobile, licensing, leaderboards and the itch.io page
A launch plan for a browser game built with an AI agent: performance budget, device tests, audio and touch fixes, an asset license ledger, leaderboard hardening and the itch.io page.