{
  "$schema": "https://deeplugin.store/data/launch-packet.schema.json",
  "version": 1,
  "plugin": {
    "id": "deeplugin-0244723c0a12f6f44919",
    "name": "oss-prompt-optimizer",
    "author": "seven282",
    "category": "tools",
    "version": null,
    "homepage": "https://github.com/seven282/oss-prompt-optimizer",
    "firstObservedAt": "2026-08-16T13:41:06Z",
    "lastChangedAt": "2026-08-19T13:11:54Z"
  },
  "links": {
    "detail": "https://deeplugin.store/plugins/deeplugin-0244723c0a12f6f44919.html",
    "feed": "https://deeplugin.store/plugins/deeplugin-0244723c0a12f6f44919.atom.xml",
    "newFeed": "https://deeplugin.store/feeds/new.atom.xml",
    "updatedFeed": "https://deeplugin.store/feeds/updated.atom.xml"
  },
  "install": {
    "spec": "oss-prompt-optimizer",
    "command": "dsh plugin --profile web add oss-prompt-optimizer",
    "agentRequest": "请先在 deeplugin.store 中检索 Registry ID deeplugin-0244723c0a12f6f44919，展示来源和精确安装 spec，为 web profile 生成安装计划；等我明确批准后再安装。"
  },
  "badge": {
    "markdown": "[![Listed on deeplugin.store](https://deeplugin.store/assets/deeplugin-listed.svg)](https://deeplugin.store/plugins/deeplugin-0244723c0a12f6f44919.html)",
    "html": "<a href=\"https://deeplugin.store/plugins/deeplugin-0244723c0a12f6f44919.html\"><img src=\"https://deeplugin.store/assets/deeplugin-listed.svg\" alt=\"Listed on deeplugin.store\"></a>"
  },
  "posts": {
    "zh": "oss-prompt-optimizer 已收录到 DeepSeek Harness Plugin Store。将原始指令经 harness llm 服务优化为专业提示词——四段（Role / Task / Context / Format）或无标题纯文本风格；情境感知（1.3.0+）：自动解析指令与对话为角色/任务/目标画像，角色与任务强相关、目标与约束自动保留（目标对齐重试、会话级目标沿用），角色按「身份＋能力＋行为」三要素撰写并附任务类型写法建议（1.3.3-1.3.5）；上下文感知默认开启；长输出断点续传+自动扩容、结果缓存（同请求零模型调用）、流式早停与可选速档降低耗时（1.3.6）；输入框 ✨ 一键优化/撤销/取消与 token 消耗提示、prompt_optimize 工具（优化+迭代）、自动优化钩子与中英自动检测的角色文档。\n\n查看来源与安装：https://deeplugin.store/plugins/deeplugin-0244723c0a12f6f44919.html",
    "en": "oss-prompt-optimizer is now listed on the DeepSeek Harness Plugin Store. Optimize a raw instruction into a professional prompt through the harness llm service — four-section (Role / Task / Context / Format) or a headerless plain-text style. Situation-aware (1.3.0+): the instruction and conversation are parsed into role / task / goal profiles, so the role stays on-task and goals/constraints are preserved (goal-alignment retry, per-session goal carry-over); roles follow an identity + capability + behavior formula with per-task-type writing tips (1.3.3-1.3.5). Context-aware by default; long outputs resume with auto-expanded budgets; validated results are cached (identical requests cost zero model calls); stream early-stop and an optional fast profile cut latency (1.3.6). Composer ✨ one-click optimize / undo / cancel with a token-cost hint; prompt_optimize tool (optimize + iterate), auto-optimize hook, auto-detected zh/en role document.\n\nSource and install details: https://deeplugin.store/plugins/deeplugin-0244723c0a12f6f44919.html"
  }
}