{
  "name": "dsh-gpt-image",
  "owner": "yyfabab1",
  "url": "https://github.com/yyfabab1/dsh-gpt-image",
  "category": [
    "image-generation",
    "video-generation",
    "audio-generation",
    "assets",
    "animation",
    "tools",
    "workflow"
  ],
  "description": {
    "en": "AI asset production pipeline for DeepSeek Harness. From a one-line prompt to files your game actually loads: generate (image / video / audio) -> machine self-check -> scale to game spec -> install into the project -> register in the asset table -> verify the code really references it. Includes a coherent multi-frame animation pipeline (anchor image -> image-to-video -> frame extraction -> alpha keying -> shared-bbox spritesheet + JSON), colour variants that inherit animation data, grid-based animation install, engine schema adapter, and a disk guard. 31 agent tools + 5 sidebar pages, zero runtime dependencies, 729 tests.",
    "zh": "DeepSeek Harness 的游戏素材生产流水线：从一句话描述到**游戏真的会加载的文件** —— 生成（图/视频/音频）→ 机械自检 → 按游戏规格缩放 → 装进项目 → 登记进素材表 → 核对代码真的引用了。含连贯多帧动画（锚点图 → 图生视频 → 抽帧 → 抠底 → 共同包围盒 → 精灵表+JSON）、继承动画数据的配色变体、按格子缩放的动画装机、引擎 schema 适配层、磁盘守卫。31 个工具 + 侧栏 5 页，零运行时依赖，729 个测试。"
  },
  "npm": "@dsh-external/dsh-gpt-image",
  "install": "dsh plugin --profile web add @dsh-external/dsh-gpt-image",
  "added": "2026-10-01",
  "capabilities": [
    "fs-read",
    "fs-write",
    "network",
    "credentials",
    "env"
  ],
  "screenshots": []
}