- Home
- Skills
- Intellectronica
- Agent Skills
- Copilot Sdk
copilot-sdk_skill
- Python
151
GitHub Stars
1
Bundled Files
2 months ago
Catalog Refreshed
4 months ago
First Indexed
Readme & install
Copy the install command, review bundled files from the catalogue, and read any extended description pulled from the listing source.
Installation
Preview and clipboard use veilstrat where the catalogue uses aiagentskills.
npx veilstrat add skill intellectronica/agent-skills --skill copilot-sdk- SKILL.md7.1 KB
Overview
This skill provides practical guidance for building agents and applications with the GitHub Copilot SDK across TypeScript, Python, Go, and .NET. It focuses on SDK usage patterns, CLI configuration, custom tools, MCP server integration, and creating custom agents. Use it to accelerate development, avoid common pitfalls, and follow recommended lifecycle and event patterns.
How this skill works
The skill explains how the SDK manages a Copilot CLI server process and talks to it over JSON-RPC via a Client → Session → Event model. It covers starting/stopping the client, creating and resuming sessions, registering event handlers before sending prompts, and handling streaming deltas and final assistant messages. It also shows how to define and register custom tools, configure external CLI servers (cliUrl/cli_url), and wire MCP servers for enhanced agent mode.
When to use it
- Creating a new agentic application that invokes Copilot programmatically
- Adding custom tools or tool handlers to a Copilot-driven workflow
- Configuring a shared MCP server or custom CLI endpoint for teams
- Migrating a prototype across TypeScript, Python, Go, or .NET
- Debugging session lifecycle, streaming events, or tool invocation issues
Best practices
- Register event handlers before calling send() to capture all events reliably
- Prefer sendAndWait or explicit session lifecycle methods for deterministic flows
- Use streaming:true when you need incremental output, but always expect the final assistant.message
- Keep CLI config in ~/.copilot/ (config.json and mcp-config.json) and document custom agent settings
- Validate tool schemas (Zod/Pydantic) so tool inputs/outputs are predictable across languages
Example use cases
- A TypeScript app that spins up a Copilot session, streams partial answers, and updates a UI in real time
- A Python microservice that exposes a lookup tool to fetch database records for agent prompts
- A Go backend that connects to a centrally hosted MCC/CLI server via cliUrl for shared capacity
- A .NET application that resumes long-running sessions and orchestrates multiple tool invocations
- Creating a custom coding agent with a tailored systemMessage and specialized developer tools
FAQ
The SDK supports TypeScript/Node.js, Python, Go, and .NET with idiomatic APIs for each.
Do I need a Copilot subscription?
Yes. A Copilot subscription (Pro, Pro+, Business, or Enterprise) and an authenticated Copilot CLI are required.