Game Jam Agent Skill

Game Jam

Turn a theme and a fixed deadline into a finished, submitted game. This is a planning and

scope-control playbook, not engine code: the win condition is *something submitted and

playable*, not *the game you imagined*.

When to use

• Use when entering a timed jam (Ludum Dare, GMTK Jam, Global Game Jam, a weekend jam),

scoping a 48-hour build, or deciding what to cut to hit a deadline.

• Use when the user asks "what can I actually build in a weekend" or "how do I submit to

this jam".

When *not* to use: building the *core loop* in engine code (use the engine skill —

godot-2d-movement, phaser-core, etc. — or a genre skill like platformer); throwaway

experiments with no deadline (use prototype-fast); shipping a commercial release (use

steam-publish / itch-publish).

Core workflow

• Read the rules before the theme drops. Confirm the jam's *length*, *theme reveal

time*, *submission deadline (with timezone)*, whether teams/pre-made assets/engines are

allowed, and whether ratings require you to rate other entries. Missing one of these

disqualifies an otherwise-finished game.

• Prep the boring parts in advance (allowed by most jams): empty project that builds

and exports, input + a title/end screen stub, an export pipeline you've run once, and a

font/SFX source you're licensed to use. Day-one time is too precious to spend here.

• Theme → one sentence. Brainstorm 10 ideas in 15 minutes, then commit to one

expressible as: *"You [verb] to [goal] while [constraint]."* If you can't say it in one

sentence, it's too big.

• Scope to the clock, not the idea. Use the budget table below. Pick one core

mechanic and one "hook". Everything else is a stretch goal.

• Build the vertical slice first. Get a 30-second playable loop (start → play → lose/win

→ restart) running early. A complete tiny loop beats a half-built big one.

• Reserve the last ~20% of the clock for shipping, not features. Export, test the build

on a clean path, capture screenshots, write the page. Builds always break at hour 47.

• Submit early, update if time remains. Upload a working build *well before* the

deadline; most jam pages let you swap the file until it closes. A submitted mediocre game

scores; an unsubmitted great one does not.

Patterns

1. Scope budget by jam length (plan backwards from the deadline)

              | 48-hour jam            | 72-hour jam           | 1-week jam
--------------|------------------------|-----------------------|------------------------
Core mechanic | 1, proven by hour 6    | 1-2                   | 2-3 interacting
Levels/content| 1 hand-made or gen'd   | 3-5                   | 8-12 + progression
Art           | placeholder/1 palette  | cohesive 1-artist set | themed set + UI
Audio         | 3-5 SFX + 1 loop       | SFX + 1-2 tracks      | adaptive music
SHIP BUFFER   | last 8-10 h            | last 12-16 h          | last full day

2. The 48-hour timeline (concrete blocks)

H0-2    Ideate -> lock ONE-sentence concept -> name the core mechanic + win/lose.
H2-8    Core loop in code with primitives (boxes/circles). Make it playable, no art.
H8-12   Sleep. (Tired code is tomorrow's bug list.)
H12-24  Content: 1 level/encounter, tune difficulty, add the "hook" feature.
H24-30  Sleep + playtest with one other person; cut anything not landing.
H30-38  Art + audio pass. Juice: screenshake, hit-stop, tween, particles, SFX.
H38-44  Bug-fix freeze: no new features. Fix only crashes + soft-locks.
H44-48  EXPORT, test the build clean, screenshots, write page, SUBMIT (by ~H46).

3. Feature triage — decide fast, decide out loud

For every feature ask, in order:
  1. Does the core loop work WITHOUT it?   -> if yes, it's a stretch goal, not MVP.
  2. Can a player tell it's missing?       -> if no, cut it.
  3. Is it < 30 min of work?               -> if no, defer past the ship buffer.
Default answer under deadline pressure is CUT. You can always add in a post-jam version.

4. Pre-submission build check (run on a clean copy, not your dev folder)

[ ] Build runs from an extracted zip on a path with NO engine/IDE installed.
[ ] Controls are shown on-screen or on the page (jurors won't read your mind).
[ ] No dev console errors; no soft-lock; restart works.
[ ] Web build (if any) is set to the platform's playable-in-browser mode.
[ ] Page has: 1-line pitch, controls, screenshots, credits + asset licenses.
[ ] File uploaded and the submission is actually attached to the JAM, not just the page.

Pitfalls

• Scope creep is the #1 jam killer. If the core loop isn't playable by the first third

of the clock, cut features now, not later.

• No ship buffer. Exporting, zipping, and writing the page reliably takes 1-3 hours and

always surfaces a build bug. Treat the deadline as ~2 hours earlier than it is.

• Submitting to the page but not the jam. On itch.io a jam entry is a separate

*submission* linked from the jam page — uploading the game alone does not enter it.

• Untested export. "Works in the editor" is not "works in the build." Test the exported

artifact on a clean path; missing assets and wrong working directories are classic.

• Unlicensed assets. Music/fonts/sprites pulled from the web can violate jam rules and

break later distribution. Use assets you're licensed for and list them in the credits.

• Polishing before the loop is fun. Juice multiplies fun; it can't create it. Get the

loop right with placeholders first.

References

• For the throwaway-vs-keep prototyping mindset and greyboxing technique, read the

prototype-fast skill.

• For the actual upload mechanics (project page, channels, butler push), read

itch-publish.

Related skills

• prototype-fast — validate a mechanic quickly before committing jam hours to it.
• itch-publish — create the page and upload the build (most jams are hosted on itch.io).
• Engine cores (phaser-core, love2d-core, godot-gdscript, …) and genre skills

(platformer, roguelike, …) — build the actual loop the jam scopes around.