Game Jam Kit: a 48-hour Godot 4 plan and a two-day Unity PRD that a coding agent can follow
by PaperChaseWebb
A tested jam method for working with a coding agent: seven milestones per engine, each with a first version, a first test and a written list of what to leave out, a JAM.md the agent cannot change without your new yes, three owner approvals (scope lock, art and audio licenses, upload), a scope-cut checklist and a fifteen-check read-only jam_check.py. Includes a playable Godot 4 starter that tests itself headless, and a tested proxy for games that call an LLM while people play.
- Price
- $29
- Difficulty
- Intermediate
- Setup
- 1 hr
- Format
- .md file
game-jam-kit-godot-48-hour-plan-and-unity-two-day-prd.md · 90 KB · instant download after checkout
Want this built for you instead? We can build it for you.
What you end up with
Your agent builds the next milestone and nothing else, the goal is on screen in the first ten seconds, every asset has a license row, and nothing gets uploaded without your yes on that exact build.
What you get
- 23 sections in one .md file
- A 48-hour Godot 4 plan and a two-day Unity 6 PRD, seven milestones each with a v1, a first test and what to leave out
- JAM.md: the plan format with times, objective, first ten seconds, cut list, AI disclosure and LLM settings
- Three owner approvals tied to hashes: scope lock, art and audio license check, and upload of a named build
- A scope-cut checklist with the order of questions for a late milestone
- jam_starter.py: a playable Godot 4 game, the plan and the ledgers in one command, with a headless self-test
- jam_check.py: fifteen read-only checks on a Godot or Unity jam folder
- A seeded fixture with 15 planted faults and a Unity fixture with 4, with the expected output
- Can a released game call an LLM? llm_proxy.py with per-player and daily caps, and llm_client.gd with a fallback line, tested together
- Rules blocks for any harness plus Godot, Unity and playtester add-ons, twelve prompts and setup for Grok Bot, Claude Code, Codex and Hermes
- Overnight run rules, the evening-before checklist, an after-jam review and a failure table
- A look at six game prompt products and kits, with sources
Who it is for
Jam teams and solo developers using a coding agent in Godot 4 or Unity 6 who have lost a jam to scope creep, an unclear goal, an asset license question or a broken web build at the last hour.
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: The kit is the only purchase. Godot is free; Unity and model usage are billed by those vendors. The LLM proxy only costs what your caps allow.
Time to result: Set up and tested the evening before the jam in about an hour
How it works
Test the check
Save the scripts from the file, build the fixture and run jam_check.py on the four fixture folders.
Expected: 15 findings on seeded, 0 on fixed, 4 on unity-seeded, 0 on unity-fixed; exit 2 on a missing folder.
Write the starter
Run python3 jam_starter.py my-jam, then the self-test with Godot headless. For Unity, copy JAM.md, the ledgers and AGENTS.md into your project.
Expected: SELFTEST PASS, and the game opens with the goal on screen.
Fill JAM.md and the cut list
Write the theme, objective, first ten seconds and AI disclosure, adjust the seven milestones and pick at least three cuts from section 8.
Expected: jam_check.py shows only J06, because the scope is not locked yet.
Lock the scope
Run jam_check.py --hash, have the agent show the scope card, and reply with your yes and the hash.
Expected: An approval row with your words, and jam_check.py prints 0 findings.
Run the jam one milestone at a time
Use the milestone prompt, export and test at each milestone, give the license yes after the art pass and the upload yes for one named build.
Expected: Every done milestone has a tested export row and nothing is uploaded without a submission row for that build.
Sample prompt from the pack
Do {milestone} only. The leave-out list is a hard limit. After each change run the self-test or batchmode tests and paste the exit code.
When the v1 exists, export, run the first test, write the builds.csv row and stop.Prompt 5 of 12. Runs after the rules block in section 14.
{milestone} is {hours} hours late. Use the scope-cut checklist questions in order and propose up to three cuts.
Do not cut anything yet; show me the cuts and the new scope hash.Prompt 7 of 12. A cut is a scope change, so it needs a new lock.
Free preview of the file
# Game Jam Kit: a 48-hour Godot 4 plan and a two-day Unity PRD that a coding agent can follow Version 1.0, 2026-10-28. By PaperChaseWebb. Works with Grok Bot, Claude Code, Codex and Hermes. Plain Markdown, three Python 3 scripts (standard library only) and two GDScript files, tested on a seeded fixture with fifteen planted faults. The Godot parts were run on Godot 4.7.2 headless. The Unity parts were not run in Unity. This kit is a way to finish a game jam with a coding agent doing most of the typing and you making the calls. You get a 48-hour Godot 4 plan and a two-day Unity 6 PRD, each cut into seven milestones. Every milestone has a first version, a first test and a written list of what to leave out. The agent works one milestone at a time from a JAM.md file it is not allowed to change. You say yes at three points: the scope lock, the art and audio license check, and the upload. A read-only checker, jam_check.py, reads the jam folder and tells you where the plan, the licenses or the approvals have slipped. It also answers a question jam builders keep asking: can a published game call an LLM while people play it? Yes, if the key stays on a server you control, the game rules stay in code and the spend has a hard cap. Section 10 has a tested proxy and a Godot client for exactly that. What it does not repeat: how to set up each engine for an agent (trap tables, headless and batchmode loops, editor scripts) is in Engine Agent Kits (pp-engine-agent-kits). Making the agent pl
Worth knowing
- Read your jam's rules on AI before you start. Some jams accept only entries made without generative AI; this kit does not help you get around that.
- The Godot plan, starter, client and checks were run on Godot 4.7.2 headless. The Unity PRD and commands were not run in Unity for this release; the Unity checks were tested on plain project files.
- The LLM proxy is a jam starting point. Put it behind HTTPS on your own host and add a moderation step before a public release.
- No claim about placing in a jam. The kit keeps the plan, licenses and approvals honest; the game is yours.
- On 2026-10-28, jam_check.py reported the fifteen expected findings on the seeded fixture, then 0 on the fixed one, 4 then 0 on the Unity fixture, with file hashes unchanged and exit 2 on a missing folder; the starter's self-test passed headless; the proxy and the Godot client returned the model's line, the capped fallback and the offline fallback. Run the tests on your own machine first.
Questions
Can a published game call an LLM at runtime?
Yes, if the key lives on a server you control, the game rules stay in code and the spend has a hard per-player and daily cap. The kit includes a tested standard-library proxy and a Godot client that shows a fixed line when the server is down or a cap is hit.
Does it work without an agent?
Yes. The plan, cut list, approvals and checks work for any team. The rules blocks and prompts are there for when an agent writes the code.
Why does the agent need my yes three times?
Scope, licenses and the upload are the three calls that cost a jam when they go wrong, and they are yours to make. Each yes is tied to a hash or a build, so a change after your yes shows up in the check.
How is this different from Engine Agent Kits?
Engine Agent Kits sets each engine up for an agent: trap tables, headless and batchmode loops, editor scripts. This kit is the clock on top of that: what to build in which hour, what to leave out, and when to stop and ask you.
Details
Version 1.0 · Updated Oct 28, 2026 · Markdown version of this page
License: Use on your own games and on installs you do for clients and teams. Do not resell or republish the kit itself. Download terms
Related
A playable Godot jam starter that tests itself (free sample)
From the Game Jam Kit: one line writes a small playable Godot 4 game with its goal on screen from the first frame, a 48-hour plan with seven milestones, ten agent rules and three empty ledgers. The game plays itself headless and exits 0 or 1, a quick read-only check covers five common jam faults, and the docs answer whether a released game can safely call an LLM.
Game Dev with AI Agents: method, prompts, skills and check loops
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.
Engine Agent Kits: Godot 4, Unity 6 and Unreal 5 rules files, editor bridges and build loops for coding agents
One kit per engine for coding agents: a rules file, a trap table of version mistakes, terminal commands to check, test and build without clicking, ways to give the agent eyes, asset pipelines with a license ledger and a one-milestone work loop.
Two-Agent Build and Test Loop for paid apps: one agent builds, another tests, and nothing passes on its own say-so
A tested setup for apps that take money: one agent writes the code, a separate tester session writes the stories and tests and gives the verdicts, a small runner records every red, green and revert exit code itself, and the owner says yes to the plan, any payment or pricing work, each merge and each release. Test keys and sandboxes only. A read-only check with fifteen tests audits the finished run.
Make your agent play the game before it says done
One prompt, two agents: Claude Code built a neon racing game far faster, while Grok Bot drove the race with real controls and caught reversed steering a script never touched.
Let AI write the NPCs but keep the rules in code
A one-person studio is building Someone Else's War, a Steam RPG where an AI writes the world, characters and replies to whatever you type, while hand-written rules decide every fight and hold NPCs to the deals they make.
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.
Write patch notes and store page copy players actually read
15 prompts that turn your commit log or task board into clear patch notes, a Steam or itch.io store page and devlog posts in your own voice.
Port a Three.js game to Unity and keep the hand-tuned physics
Moving a browser racing game to Unity for phones, the developer asked Claude Code which Asset Store engine to use. It read the code and advised using neither, then ported the 170 lines of drift physics line by line so the feel stayed the same. The port and a new UI took about two hours.
Game Assets as Code: one spec file, seeded generators and a fifteen-check checker for agent-made game art and sound
Let Grok Bot, Claude Code, Codex or Hermes make your placeholder and jam-ready game assets as code instead of prompts: pixel sprites from letter grids, seeded symmetric sprite variants, seamless tiles, SVG icons, retro sound effects and low-poly Blender props built headless, all from one spec file with a fixed palette and seeds. Every build writes a manifest with file and spec hashes and a review page; your approvals are tied to those hashes, a promote step copies only approved, unchanged files into the game, and a separate release yes covers anything that leaves your machine.