Impart Mcp — independently scanned and version-tracked by SaferSkills.
SaferSkills independently audited Impart Mcp (Agent Skill) and scored it 100/100 (green). The audit ran 55 deterministic rules across Security, Supply Chain, Maintenance, Transparency, and Community; it found 0 high-severity and 0 lower-severity findings. The full rule-by-rule trace and per-finding evidence are below. Free, methodology-open.
Findings & checks · 0 flagged
Every scanned point with the score it earned and what moved between them.
First recorded scan — no prior version to compare against.
The primary manifest — the file an agent reads to learn what this artifact does.
impart-mcp 是一个基于 Model Context Protocol (MCP) 的智能体编排服务器。它充当「Agent 编排层」,将具备高度工程化能力的专家智能体(如 Claude Code, Codex CLI 等)封装为标准 MCP 工具,使宿主环境(Claude Desktop, Cursor 等)能够通过统一接口调用这些具备复杂任务处理能力的「超级 Agent」。
| Agent 名称 | 核心定位 | 默认模型 | 权限 |
|---|---|---|---|
| Advisor | 架构与调试专家,处理高难度工程问题 | codex/gpt-5.2 | Read-only |
| Researcher | 代码研究专家,擅长跨仓搜索与文档分析 | claude/sonnet | Read-only |
| Explore | 代码定位专家,快速回答「某逻辑在哪里」 | gemini/gemini-3-flash-preview | Read-only |
| Frontend Engineer | UI/UX 实现专家,独立完成高保真组件 | gemini/gemini-3-pro-preview | Read-write |
| Document Writer | 技术文档专家,撰写清晰的规范文档 | gemini/gemini-3-flash-preview | Read-write |
无需安装,直接在 MCP 配置中使用:
npx impart-mcp@latestnpm install -g impart-mcp
# 或
bun add -g impart-mcpgit clone <repository-url>
cd impart-mcp
bun install # 或 npm install
bun run build # 构建产物impart-mcpcommandnpx impart-mcp@latest(或全局安装后直接使用 impart-mcp)使用 claude mcp add 命令一键添加:
claude mcp add impart-mcp -- npx impart-mcp@latest或全局安装后:
claude mcp add impart-mcp -- impart-mcp默认情况下,所有 Agent 使用内置的默认配置即可运行,无需额外配置。
如需自定义 Agent 的模型或开关,可通过环境变量 AGENT_CONFIG_PATH 指定配置文件路径:
{
"mcpServers": {
"impart-mcp": {
"command": "npx",
"args": ["impart-mcp@latest"],
"env": {
"AGENT_CONFIG_PATH": "/path/to/your/agent-config.json"
}
}
}
}自动初始化:若指定的配置文件不存在,系统会在首次运行时自动生成包含所有默认值的配置模板,方便你按需修改。
配置文件示例 (`agent-config.json`):
{
"advisor": {
"model": "codex/gpt-5.2",
"enabled": true
},
"researcher": {
"model": "claude/sonnet",
"enabled": true
},
"frontend-ui-ux-engineer": {
"model": "gemini/gemini-3-pro-preview",
"enabled": true
}
}服务器提供了以下核心工具:
| 工具名称 | 类型 | 描述 |
|---|---|---|
call_agent | 同步 | 阻塞式调用,适用于需要立即验证结果的场景 |
call_agent_async | 异步 | 后台启动任务,返回 task_id,推荐用于耗时较长的搜索任务 |
get_agent_result | 查询 | 获取异步任务状态,支持 block 模式阻塞等待 |
call_agents_batch | 批量 | 并行调用多个 Agent,阻塞直到所有任务完成 |
`call_agent` / `call_agent_async`:
agent:Agent 名称枚举 (advisor | researcher | explore | frontend-ui-ux-engineer | document-writer)prompt:任务描述cwd:工作目录(必填)context:可选上下文信息| 命令 | 描述 |
|---|---|
bun start | 启动服务(Bun 运行时) |
bun dev | 开发模式(支持热重载) |
bun run build | 构建产物到 dist/ |
node dist/server.js | 启动服务(Node.js 运行时) |
| 文件 | 职责 |
|---|---|
src/server.ts | MCP 协议处理与工具注册 |
src/agents/ | Agent 的 System Prompt 及行为定义 |
src/providers/ | 模型供应商适配逻辑 |
src/types.ts | 全局类型定义 |
~30 seconds. Free. No account. Every finding cites a rule and a line of evidence.