tether-ai
AGENTS.md92 starsChanged 55 days agoMITView on GitHub
面向仓库的 AI 编程桌面工作台。Agent 循环、沙箱、会话在 npm 包 tether-agent-core;本仓库只做 Electron 壳和中文工作流界面。 Skills 由 Pi 运行时加载(Tether 不另写 loader)。标准路径: 每个 skill 目录一个 SKILL.md,frontmatter 需含 name 与 description(Pi 校验,缺项不会加载)。 macOS / Linux: Windows PowerShell:
In this repository5 other files it gives its agents.
- continue-long-runSkill
- init-long-runSkill
- plan-then-actSkill
- tether-uiSkill
- tether-uiSkill
The file
Files
Tether
面向仓库的 AI 编程桌面工作台。Agent 循环、沙箱、会话在 npm 包 tether-agent-core;本仓库只做 Electron 壳和中文工作流界面。
地图
- 产品说明与架构:README.md
- 主进程(窗口 / IPC / 权限 / 工作区):src/main/index.ts
- RPC 子进程宿主:src/main/agent-host.ts
- 界面状态:src/renderer/App.tsx
- 组件:src/renderer/ui.tsx
- 会话归并:src/renderer/conversation.ts
- IPC 契约:src/shared/types.ts
- 长任务协议:
.agents/skills/init-long-run、.agents/skills/continue-long-run、.agents/skills/plan-then-act - 界面气质(改 CSS / 组件 / 克隆同款工作台):
.agents/skills/tether-ui(Cursor 侧同步.cursor/skills/tether-ui)
Agent Skills
Skills 由 Pi 运行时加载(Tether 不另写 loader)。标准路径:
| 范围 | 路径 |
| 项目(需信任) | .agents/skills/<name>/SKILL.md、.pi/skills/<name>/SKILL.md |
| 用户全局 | ~/.tether/skills/<name>/SKILL.md、~/.agents/skills/<name>/SKILL.md |
每个 skill 目录一个 SKILL.md,frontmatter 需含 name 与 description(Pi 校验,缺项不会加载)。
- 调用:输入
/skill:名称;输入/时也会列出当前会话已加载的 skill - 查看:设置 → Agent Skills(展示路径与已加载列表)
- 项目 skill 需先信任项目;
@文件引用仅扫描项目内.agents/skills与.pi/skills
约定
- 新对话必须先选项目;绑定 cwd,发第一条消息才
startAgent。 - 生成中回车进入排队(最多 5 条,不入会话文件);当前回合正常结束后自动发出队首。停止或出错会暂停队列,界面上可以继续发送或对单条立即插话(运行中走
steer)。换对话/项目清空。/命令不进队。 - 权限:
plan/ask/auto/full。有项目时sandbox: workspace-write。 - 跨会话进度写在
.agents/features.json和.agents/progress.md,不要只写在聊天里。 features.json只改passes,不要删条目或改description。- 面向用户的文案用简体中文;代码、路径、命令保持原文。
检查
macOS / Linux:
pnpm test
pnpm typecheckWindows PowerShell:
pnpm test
pnpm typecheckDoes it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

