by alexanderizh
SparkWork is a local-first desktop Agent workbench for professional development and complex work. Orchestrate workflows, collaborate across Agent sessions, route tasks to the right models, and extend the workspace with custom tools and apps. Reviewable, steerable, reversible runs on macOS, Windows and Linux.
# Add to your Claude Code skills
git clone https://github.com/alexanderizh/spark-agentSee how spark-agent compares with popular alternatives.
spark-agent is an open-source ai agents skill for AI coding assistants such as Claude Code, Codex CLI, and ChatGPT, built by alexanderizh. SparkWork is a local-first desktop Agent workbench for professional development and complex work. Orchestrate workflows, collaborate across Agent sessions, route tasks to the right models, and extend the workspace with custom tools and apps. Reviewable, steerable, reversible runs on macOS, Windows and Linux. It has 104 GitHub stars.
spark-agent's catalog security scan is still queued. You can run an instant dependency and prompt-injection check now with the "Scan for vulnerabilities" button above.
Clone the repository with "git clone https://github.com/alexanderizh/spark-agent" and add it to your Claude Code skills directory (see the Installation section above).
spark-agent is primarily written in TypeScript. It is open-source under alexanderizh on GitHub, so you can review or fork the full source.
Yes. SkillsLLM lists many other AI Agents skills you can browse and compare side by side. Open the AI Agents category from the badge at the top of this page, or use the Related Skills and comparison links further down to weigh spark-agent against similar tools.
No comments yet. Be the first to share your thoughts!
⚠️ Third-Party Software Notice
This skill is third-party open-source software developed and hosted independently on GitHub. SkillsLLM is an informational directory and does not control or maintain the underlying repository.
Any security checks, ratings, or warnings displayed by SkillsLLM are automated and limited in scope. They do not constitute a security certification or guarantee that the software is safe, error-free, or free from malicious code, vulnerabilities, compromised dependencies, or prompt-injection risks.
Review the source code, permissions, dependencies, and configuration before installing or running any third-party skill. Use is at your own risk. To the maximum extent permitted by applicable law, SkillsLLM is not liable for losses arising from third-party software.
The deep catalog scan for this skill is still queued. Run an instant dependency check now instead.
本地优先的桌面 AI Agent 工作台:在一个应用里完成对话、开发、调研、文档、多媒体创作与自动化。
官网 · 下载 · 能力图解 · 功能特性 · 快速开始 · 从源码构建 · 工程手册 · 贡献指南 · 更新日志
SparkWork(仓库名 Spark Agent)是一个运行在 macOS、Windows 和 Linux 上的桌面 AI Agent 工作台。
项目仍在快速迭代,部分功能和界面可能继续调整。

下面八张图对应八条能力线的主链路;逐项细节见官网文档中心与工程手册。
长期记忆:一轮对话的写入与读取闭环
每轮结束异步抽取写入、每轮开始混合检索注入;写入过五道闸门,读取走 FTS5 与向量双路、RRF 融合后按 token 预算裁剪。
知识库:从生成到读取的架构
三条生成管道(对话沉淀 / 技能提议 / Repo Wiki)都先出草案、再人工确认,写入收敛到同一原语;读取零预注入,人机同源,先检索、再按需取正文。
工作流:一张图的真实执行
13 种节点类型经拓扑调度逐个真实派发:agent 节点是同一会话内的成员 turn,工具节点可配置成不经 LLM 的确定性直调;快照落库支持断点续跑。
自定义工具:从工程到 Agent 工具面
多文件工程安装为不可变版本,密钥只进系统凭据库;过完整门禁启用后挂进会话工具面,包变更下一轮热生效。
自动路由:一轮分流与降级阶梯
分流器逐轮判强度,宿主在轮次开始前替换 provider 与 model;分流失败一律降级,不会中断本轮执行。
性能保护:采样、定级与四档滞回
2 秒采样六类指标、逐项定级后取最高档,升级快降级慢;只有危急档才弹窗提醒,前两档只在性能页可见。
安全与密钥:一枚密钥的受控旅程
密钥明文直接被拒,只走受保护表单进系统凭据库;环境变量按六级作用域合并注入子进程,hooks 分观察与可拦截双轨,全程日志脱敏。
上下文工程:长上下文的节流与可信
技能目录只注入元数据、全文按需加载;超长工具输出归档为内容寻址制品、按需读回;压缩只追加事件、档案无损可检;跨会话只读引用须显式授权。
记忆生命周期、知识库沉淀管道、router 组成与双池闸门这四张图在工程手册 · 能力架构里。



与 Agent 协作的完整闭环:从输入、排队、审批到追问、分叉、定时唤醒,长任务不丢线索。
/ 即达对话右侧是一套可多开、可切换的面板体系——任务的每个侧面都能边看边聊。



在 GitHub Releases 获取最新安装包:
| 平台 | 格式 |
|---|---|
| macOS | .dmg |
| Windows | .exe / NSIS 安装包 |
| Linux | .AppImage / .deb |
桌面应用已内置 spark-engine(SDK 进程内运行),日常使用无需额外安装 CLI。
仅当需要在终端里使用 spark 命令 / TUI 时才单独安装;CLI 与桌面应用共用 ~/.spark 数据根,可复用桌面端已配置的渠道凭据。
macOS / Linux(要求 Node.js >=22.14 <23):
curl -fsSL https://raw.githubusercontent.com/alexanderizh/spark-agent/spark-cli-releases/install.sh | sh
Windows(PowerShell):
irm https://raw.githubusercontent.com/alexanderizh/spark-agent/spark-cli-releases/install.ps1 | iex
安装脚本会下载对应版本的 npm tarball 并对照 latest.json 校验 sha256,支持指定版本与离线安装:
sh install.sh --version 0.4.0 # 安装固定版本
sh install.sh --tarball ./spark-agent-0.4.0.tgz # 本地 tarball 离线安装
更新(事务式自更新,失败自动回滚):
spark update # 检查并更新到最新版
spark update --check # 仅检查是否有新版本
spark update --target 0.4.0 # 安装 / 回退到指定版本
spark update --allow-prerelease # 允许预发布版本
安装或更新后运行 spark --version 与 spark doctor 验证环境。
首次启动会在选择器里选一次模型,之后选择会写入 ~/.spark/config.toml([agent].model)自动复用,不再每次询问;切换用 /model 或 spark --model <route-id>(仅当次生效)。
会话管理:会话按目录持久化在 ~/.spark/projects/ 下,可随时恢复:
spark --continue # 继续当前目录最近一次会话(-c)
spark --resume # 打开会话选择器(TUI 内也可用 /sessions 切换)
spark --resume <session-id> # 恢复指定会话
spark sessions # 列出当前目录的历史会话
>=22.14.0 <23>=11.13.0 <12Windows 从源码安装时建议准备 Visual Studio Build Tools,以便构建 better-sqlite3、keytar 等原生依赖。
git clone https://github.com/alexanderizh/spark-agent.git
cd spark-agent
pnpm install
pnpm dev
常用检查与打包:
pnpm typecheck # 类型检查
pnpm lint # 代码检查
pnpm test # 单元测试
pnpm build # 构建
pnpm --filter @spark/desktop build:mac # macOS 打包
pnpm --filter @spark/desktop build:win # Windows 打包
pnpm --filter @spark/desktop build:linux # Linux 打包
spark-agent/
├── apps/
│ └── desktop/ # Electron 桌面应用(主进程 + React 渲染端)
├── assets/
│ └── showcase/ # README 界面截图素材
├── packages/
│ ├── agent-runtime/ # 会话运行时与服务编排
│ ├── protocol/ # 跨进程 IPC 协议与类型
│ ├── storage/ # SQLite 存储层
│ ├── shared/ # 跨进程共享工具
│ ├── ui-kit/ # 共享 UI 组件
│ └── plugin-sdk/ # 插件 SDK
└── spark-engine/ # 终端交互子项目
主要技术栈:Electron 43 · React 19 · TypeScript · Tailwi