算法艺术创作
algorithmic-art
通过代码和可调参数创作生成艺术与交互式视觉效果。
game-dev【设计与创意】Build, extend, and verify playable browser games, optionally using SandBase for generated concept art and game assets. Use for substantial 2D or 3D web-game work; use a general frontend workflow for non-game interactive sites.
独立 GitHub 快照 · 自动分类 · 以原作者文档为准
Turn a game brief into a playable, visually coherent browser game. Keep the game code, deterministic asset processing, builds, and tests local. Use SandBase only for optional generative assets that the user requests.
Read the SandBase API map before generating concept art, textures, sprites, backgrounds, or other AI media.
Before implementation, resolve only the choices that materially affect the architecture:
Inspect the existing project and preserve its framework, package manager, engine, and conventions. For a new project, choose the smallest suitable stack: native canvas or a focused 2D engine for simple 2D games, and a WebGL engine such as Babylon.js for substantial 3D work. Do not replace an established engine merely because another is familiar.
Write a compact implementation plan for work that spans more than one mechanic. Identify the riskiest slice first: physics feel, procedural generation, animation state, navigation, camera behavior, pointer lock, shaders, networking, asset import, or performance.
Define observable acceptance criteria for each slice. A criterion should describe something a player can see or do, not merely a class or file that exists.
Keep gameplay state and rules independent from the UI shell where practical. Separate input, simulation, rendering, audio, persistence, and scene transitions enough that each can be tested without rebuilding the entire game.
Implement the smallest loop that demonstrates the intended experience:
Use temporary local primitives when necessary to prove mechanics, but do not present placeholders as finished art. Add a deterministic demo or equivalent mode when it helps visual regression checks and screenshots reach meaningful gameplay states.
Handle lifecycle boundaries explicitly: initialize the engine once, dispose listeners and render resources, resize safely, pause or mute when appropriate, and avoid duplicate loops during development remounts.
Do not call an image model merely because the task is a game. First decide whether supplied assets, procedural graphics, vector drawing, or local transformations already satisfy the brief.
When generated art is needed:
sandbase_discover with the required capability, such as text to image, image editing, or transparent background image.sandbase_inspect for viable candidates and compare the live schema, aspect ratios, reference-input support, output format, price, and execution mode.sandbase_account before an approved multi-call batch, then call sandbase_run with only current schema-defined arguments.sandbase_run_get using the same run ID. Never create a duplicate because a run remains pending.sandbase_runs only to recover status or reconcile actual cost.Record each generated asset's purpose, prompt summary, endpoint, run ID, local path, dimensions, provenance, and observed cost. Verify that sprites and textures work at actual gameplay scale; attractive source images are not automatically usable game assets.
Do not imitate a living artist's identity or use protected characters, logos, faces, or voices without the necessary rights. Send only the minimum user material required by the inspected endpoint.
Treat game feel as part of correctness. Tune acceleration, friction, camera lag, hit feedback, animation timing, sound cues, and difficulty against the stated experience rather than arbitrary constants.
Run the project's type checks, tests, and production build. Then exercise the game in a real browser at representative viewport sizes and inputs.
Verify at least:
When code and the rendered result disagree, treat the rendered result as the defect report. Do not claim completion while the core loop is unplayable, generated assets are still pending, or the production build fails.
Return the playable project or requested build, control summary, implemented mechanics, validation results, asset ledger, SandBase run IDs and costs when applicable, and any known limitations. Deployment or publication is a separate external action unless the user explicitly requested it.