What does this skill do, and when should you use it?
This skill helps developers build applications that interact programmatically with GitHub Copilot through the Copilot SDK and CLI. It covers Node.js/TypeScript, Python, Go, and .NET, including sessions, streaming, custom tools, hooks, MCP, BYOK, persistence, custom agents, and deployment. It is suited to teams embedding Copilot capabilities into backend services, desktop applications, or automated workflows. A GitHub Copilot CLI installation and authentication are normally required, along with a Copilot subscription unless BYOK is used.
Provides installation and client/session examples for four languages; explains stdio and TCP communication with the Copilot CLI; configures streaming events, custom tools, permission handlers, lifecycle hooks, and error handling; connects local or HTTP MCP servers; configures GitHub credentials or BYOK providers including OpenAI, Azure, Anthropic, and Ollama; creates, resumes, lists, and deletes sessions; and covers custom agents, skill directories, infinite sessions, and Docker Compose deployment.
- A Node.js, Python, Go, or .NET team adding programmable Copilot sessions to a backend service.
- An interactive application that needs to display model output incrementally as it arrives.
- A development team exposing business APIs, weather data, or other functions as custom tools for Copilot.
- An application that must approve tool calls, modify arguments, redact results, or record session metrics.
- An agent application that needs local or remote MCP server integrations.
- A multi-user service that must isolate sessions or resume conversations after an application restart.
How do you install this skill?
- Before using remote MCP, shell access, or persistent sessions, restrict tool sets, isolate working directories, review token propagation, and obtain per-action user confirmation; do not use tools:["*"] or environment-token examples directly in production.
- The examples depend on GitHub Copilot CLI, a subscription, or external model endpoints; verify current SDK versions, regional network reachability, authentication permissions, and language-specific API compatibility first.
- The repository is marked work in progress, and the supplied SKILL.md is incomplete at Client Configuration; documentation and CI presence should not be treated as proof of production readiness.
- Shell / CLI
- Network access
- Local filesystem
- MCP Server
GitHub Copilot CLIGitHub Copilot subscription unless using BYOKNode.js 18+Python 3.8+Go 1.21+.NET 8.0+
Install the repository's skill collection in Copilot CLI with npx skills add microsoft/skills. Select the required skill in the wizard; for GitHub Copilot, it is installed under the selected agent directory, such as .github/skills/. Install the SDK package for your language with npm install @github/copilot-sdk, pip install github-copilot-sdk, go get github.com/github/copilot-sdk/go, or dotnet add package GitHub.Copilot.SDK. The Copilot CLI must also be installed and authenticated before using the SDK.
tmp="$(mktemp -d)"
git clone --depth 1 https://github.com/microsoft/skills.git "$tmp"
mkdir -p ~/.claude/skills
cp -R "$tmp/.github/skills/copilot-sdk" ~/.claude/skills/
rm -rf "$tmp"Generated from the source repository and skill path; it copies only this skill's folder. If the author's install steps above differ, follow those first. To scope it to one project, replace ~/.claude/skills with that project's .claude/skills.
How do you use this skill?
Once installed, send your agent any of these to trigger it:
- Build a resumable Python Copilot SDK session with one custom tool.
In an Agent Skills-compatible coding agent, trigger the skill for a task such as: “Build a resumable Python Copilot SDK session with one custom tool.” Create a CopilotClient, create a session, call send or sendAndWait, and configure streaming, tools, hooks, mcpServers, provider, or sessionId as needed. Stop the client when finished. Use the language-specific examples in the skill for exact API shapes.
What are this skill's strengths and limitations?
- Supports Node.js/TypeScript, Python, Go, and .NET.
- Covers sessions, streaming, custom tools, hooks, MCP, BYOK, and deployment patterns in one guide.
- Documents both local stdio and TCP CLI connectivity.
- Includes practical guidance for persistence, permissions, error handling, and production deployment.
- Requires the GitHub Copilot CLI; an Individual, Business, or Enterprise subscription is normally also required.
- BYOK tokens do not refresh automatically, and provider configuration must be supplied again when resuming a session.
- The SDK does not provide interactive CLI UI, slash commands, login/logout flows, session export, or several other CLI-only features.
- The source does not provide an independent test suite for this skill or verification results for each runtime platform.
How does this skill compare with similar options?
Side by side with related skills; every score comes from the same FSRS standard.
| Skill | FS score | Stars | Last updated | License |
|---|---|---|---|---|
| GitHub Copilot SDK Builder this page ✓ Microsoft · Official | 49 · Use with care | ★ 2.7k | 3mo ago | MIT |
| Contracts Engine Verifier ✓ Anthropic · Official | 44 · Not recommended | ★ 425 | 1mo ago | — |
| Microsoft Agent Skills ✓ Microsoft · Official | Not yet reviewed | ★ 2.7k | 3mo ago | MIT |
| Azure AI Foundry Agent Framework ✓ Microsoft · Official | 50 · Use with care | ★ 2.7k | 3mo ago | MIT |
| Chronicle Copilot Session Analyst ✓ Microsoft · Official | 39 · Not recommended | ★ 194k | 11d ago | MIT |
Compared with using the Copilot CLI directly, this skill focuses on embedding Copilot into applications through programmatic session, message, tool, event, and deployment APIs. The CLI still provides interactive UI, slash commands, authentication flows, and session export that are not available in the SDK.
How did FollowSkills review this skill?
The document discloses stdio/TCP, MCP, model providers, tokens, session persistence, and tool execution, and includes permission-hook examples, a denial example, session deletion, and cleanup guidance. However, the remote MCP example uses bearer credentials with tools:["*"], persistence stores full conversations and tool results, and user confirmation, least privilege, sensitive-data boundaries, and rollback are incomplete, so points are deducted.
The skill provides multi-language core flows, timeout and retry behavior, error hooks, and abnormal-case guidance; the repository also contains related tests and evaluation workflows. Static review cannot verify dependency versions, API correctness, cross-language key paths, or runtime failure behavior, and the supplied skill text is truncated at Client Configuration, so the score remains below the static ceiling.
Audience, supported languages, prerequisites, core scenarios, and deployment modes are reasonably clear. Non-fit boundaries, trigger precision, version compatibility, and Chinese-language guidance are limited; core use depends on GitHub Copilot CLI, subscriptions, or remote endpoints, with no mainland-China reachability guidance, so points are deducted.
Information is organized progressively across prerequisites, installation, architecture, examples, permissions, security, persistence, and deployment. MIT licensing, official provenance, a README catalog, and security-reporting guidance are present. However, the repository declares work-in-progress status, while skill versioning, changelog, maintenance ownership, update path, FAQ coverage, and complete documentation boundaries are weak; the SKILL.md also appears incomplete.
The skill covers clients, sessions, streaming, tools, hooks, MCP, BYOK, persistence, agents, and deployment, with directly reusable examples in four languages. Correctness, current SDK compatibility, complete outputs, and comparative benefit are not statically demonstrated, so effectiveness is capped at 7.
The evidence includes the selected skill, MIT license, README, CI evaluation workflows, test code, and error-classification tests, providing some auditability. There are no executed results, coverage reports, or third-party corroboration for this skill, and no reproduction was performed, so the static ceiling of 5 applies.
Open a dimension to read why it scored that way
Evidence confidence：Low — Mostly static review, author material or a limited demo; useful for discovery, not high-risk decisions.
See the full review method →