gamedev-skills/awesome-gamedev-agent-skills / skills/ebitengine/ebitengine-game
ebitengine-game
VerifiedCreates, debugs and packages 2D games in Go with Ebitengine v2: a play-and-retry loop, input, pixel-art scaling, text, audio and desktop, WASM or mobile builds. Use when go.mod requires ebiten/v2 or the user wants a game in Go, or on blurry pixel art, missed key presses or stutter. Not for engine patches; pair with `game-feel`, `game-ui-ux` or a genre skill for design.
references/playtest-diagnostics.md
View on GitHub# Observe interactions, not just screenshots
Use when automating play, reproducing a bug or checking an important multi-step
feature. The idea is a short sequence of actions, controlled advancement,
observation and a focused change, adapted from
[develop-web-game](https://github.com/openai/skills/blob/30444aed500c00c85294d12074f6e3ee794f808a/skills/.curated/develop-web-game/SKILL.md).
Its JavaScript hooks and Playwright client are not Ebitengine APIs.
## Define a scenario with intermediate outcomes
Start from a known state/seed. Use the game's action dispatcher, with explicit
press/hold/release and tick counts. Observe intermediate states before issuing
the next action, so a successful final score does not hide a broken transition.
| Example chain | Observe along the way |
| --- | --- |
| Collect key → open door → reach exit | inventory changes once, locked/unlocked feedback agrees, next level starts with intended state |
| Hit → damage → defeat → reward | one contact event, correct health, target stops acting, one reward |
| Pause → settings → back → resume | underlying play frozen as intended, controls owned by active UI, focus restored, no leaked confirm/action |
| Invalid puzzle move → undo/reset → solve → replay | refusal explained, state/history restored, completion reachable, replay has no stale result |
Choose chains relevant to the requested change. Reset between independent
scenarios: world, RNG, timers, buffered input, pending scene transitions and
cosmetic events may all matter. Reuse the same scenario after the fix; add a
different boundary case only when it addresses an unresolved concern.
## Pair rendered evidence with actual state
For a scripted failure, a compact read-only diagnostic snapshot can include:
game/content version, seed, tick, scene/mode, logical coordinate origin and units,
camera/viewport, relevant positions, resources, timers, active UI/focus and the
last action/event. Derive it from real state; do not build a second simulation
or print expected answers. Keep it local/debug-only, concise, and free of
credentials or other players' secret state.
Capture the state and image for the **same tick/scene**. On a catch-up frame,
logging in every `Update` and later capturing in `Draw` can pair different
moments. Label the drawn capture with its state tick; report if the requested
tick was not drawn. Compare both: a JSON health value can be correct while the
visible meter is stale, and a correct-looking frame can hide a rule failure.
The starter supplies `-demo` and `-shot`; it does **not** supply a state-JSON or
general input-script flag. Add a small project-specific diagnostic when needed,
or use the pinned engine's opt-in VM tooling described in
the architecture-and-testing reference. A scripted action dispatch
tests rules/flow; actual device or VM input also tests the device-to-action path.
Run graphics checks under Xvfb on Linux, not the user's desktop.
## Control time without changing the game being checked
Advance the real fixed-step simulation rather than sleeping for an estimated
number of updates. Keep test stepping separate from the normal scheduling loop
so both cannot advance the same state. A capture-only `SyncWithFPS` mode may
accelerate tick-based rules, but does not measure interactive frame pacing,
audio synchronization or wall-clock services. Never transplant a browser
`advanceTime` hook by calling `Game.Update` concurrently with `RunGame`.
Inspect runtime errors as well as images/state. Fix the first new causal failure
before collecting downstream symptoms. Record the initial state, action/tick
sequence, changed behavior and inspected artifacts. Test only the affected
paths plus relevant regression boundaries; do not rebuild every platform after
each text or color adjustment without a reason.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| brainstormingobra/superpowersYou MUST use this before any creative work - creating features, building components, adding functionality, modifying behavior, or planning anything new, in software or out of it (a talk, a business, a renovation). | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| brand-guidelinesanthropics/skillsApplies Anthropic's official brand colors and typography to any sort of artifact that may benefit from having Anthropic's look-and-feel. Use it when brand colors or style guidelines, visual formatting, or company design standards apply. | Stars: 198k | Updated | Claude CodeCodex | Verified | Install |
| frontend-designanthropics/skillsGuidance for distinctive, intentional visual design when building new UI or reshaping an existing one. Helps with aesthetic direction, typography, and making choices that don't read as templated defaults. | Stars: 198k | Updated | Claude CodeCodex | Verified | Install |
| mcp-builderanthropics/skillsGuide for creating high-quality MCP (Model Context Protocol) servers that enable LLMs to interact with external services through well-designed tools. Use when building MCP servers to integrate external APIs or services, whether in Python (FastMCP) or Node/TypeScript (MCP SDK). | Stars: 198k | Updated | Claude CodeCodex | Verified | Install |