Agent Skills
› KunAgent/Kun
› integration-binding
integration-binding
GitHub通过显式验证接口将任务与MCP工具、工作流及资源连接。提供发现、描述、调用及资源读取等路由，规范从能力识别到结果验证的完整流程，确保权限最小化及变更可追溯，严禁虚构连接器。
Trigger Scenarios
需要调用外部MCP工具或服务
需要访问或操作外部资源文件
构建涉及第三方集成的工作流
Install
npx skills add KunAgent/Kun --skill integration-binding -g -y
SKILL.md
Frontmatter
{
"name": "integration-binding",
"description": "Connect a task to MCP tools, workflows, resources, files, or visual outputs through explicit validated interfaces."
}
Integration Binding
Copyright (c) 2026 KunAgent. Licensed under the MIT License.
Purpose
Connect a task to MCP tools, workflows, resources, files, or visual outputs through explicit validated interfaces.
Tool routing
| Tool or skill | Use |
|---|---|
mcp_search |
Discover an integration by intent. |
mcp_describe |
Load the exact input schema. |
mcp_call |
Invoke a connected MCP tool. |
mcp_list_resources |
List server resources. |
mcp_read_resource |
Read one selected resource. |
mcp_read_only_call |
Use an approved read-only integration in Plan mode. |
Workflow
- Name the required capability and whether it reads or mutates external state.
- Discover candidates instead of guessing an id.
- Inspect the selected schema.
- Call with the minimum arguments and privilege.
- Validate the returned data at the next system boundary.
Completion gates
- Use exact tool ids, resource URIs, and schemas from discovery output.
- For mutations, confirm that user authorization covers the exact external effect.
- Read back or otherwise verify externally visible changes.
Boundaries
- Never invent a connector or silently fall back to an unrelated service.
- Never include secrets or private source material in discovery queries.
Delivery
Lead with the outcome, name the evidence used for verification, and disclose any real limitation that remains.
Version History
- a984912 Current 2026-09-09 01:50
