Skip to content

README.agents.md

Latest commit

 

History

History
56 lines (40 loc) · 11.3 KB

README.agents.md

File metadata and controls

56 lines (40 loc) · 11.3 KB

Custom GitHub Copilot Agents

Custom Agents transform GitHub Copilot into a specialized expert for your specific needs. By defining a unique persona, tailored behaviors, and access to custom tools, these agents provide focused, context-aware assistance for any development workflow.

Some advanced agents are designed to connect to Managed Copilot Personalities (MCP) servers, which provide enhanced, server-side capabilities. Please check the documentation for each agent to see if an MCP server is required.

How to Use Custom Agents

Follow these steps to install and activate a custom agent in a compatible editor like VS Code.

1. Prerequisites (For MCP-Enabled Agents)

If the agent you wish to use requires a specific MCP server, you must first configure your IDE to connect to it. This is a mandatory step for these agents to function correctly.

You can typically do this in one of two ways:

  • Update mcp.json: Add the server's endpoint to your local mcp.json configuration file.
  • IDE Settings: Configure the MCP server endpoint directly within your editor's settings (e.g., in VS Code's settings.json).

2. Installation

  1. Download the File: From this repository, download the .agent.md file you wish to install.
  2. Place the File: Move the downloaded file into one of the following directories. Your editor will automatically detect it.

    • For a specific project: Place it in your project's .github/copilot/agents/ directory.

      your-project/
└── .github/
    └── copilot/
        └── agents/
            └── my-custom-agent.agent.md

    • For global use (all projects): Place it in your user-level configuration directory (e.g., ~/.github/copilot/agents/ on macOS/Linux).

3. Activation and Usage

  1. Open Copilot Chat: In your editor, open the GitHub Copilot Chat view.
  2. Select the Agent: Click on the agent selector (it may say "Default" or show another active agent).
  3. Choose Your Agent: Select your newly installed agent from the dropdown list.

Once selected, Copilot will adopt the persona and capabilities defined in the file, ready to assist you with your specialized tasks.

Title Install Description
Principal Software Architect Install in VS Code Install in VS Code Insiders A principal software architect persona that shapes resilient systems and aligns technology with strategy.
Debugging Specialist Install in VS Code Install in VS Code Insiders A diagnostic-first chat mode that hunts down defects systematically.
Senior Software Engineer Install in VS Code Install in VS Code Insiders A senior software engineer persona that designs and ships production-quality solutions with rigor.
Figma Design Orchestrator Install in VS Code Install in VS Code Insiders A Figma-native design partner that orchestrates workflows via the Figma MCP server.
Laravel Boost Delivery Lead Install in VS Code Install in VS Code Insiders A Laravel-first chat mode that ships production-ready updates via the Laravel Boost MCP server.
Learning Navigator Install in VS Code Install in VS Code Insiders A learning-first chat mode that converts curiosity into durable, shareable expertise.
Planning Strategist Install in VS Code Install in VS Code Insiders A planning-first chat mode that converts ambiguous goals into actionable, risk-aware roadmaps.
Playwright Test Orchestrator Install in VS Code Install in VS Code Insiders A Playwright-first chat mode that drives resilient E2E coverage with MCP automation.
Refactoring Strategist Install in VS Code Install in VS Code Insiders A refactoring-first chat mode that modernizes code safely and incrementally.
Security Sentinel Install in VS Code Install in VS Code Insiders A security-first chat mode that embeds OWASP-grade safeguards into every change.
Hyperfocus Sequential Strategist Install in VS Code Install in VS Code Insiders A hyperfocused Copilot persona that orchestrates the sequential thinking MCP server for disciplined, multi-step reasoning.