Cogni AI Agentic Collections: Custom Agents and Skill Sets for GitHub Copilot

This article introduces the Cogni AI Agentic Collections project, a collection of AI agents, instructions, skills, hooks, and workflow plugins designed for GitHub Copilot, supporting Claude Code. It aims to help developers customize their AI-assisted programming experience. The project offers multiple installation methods, covering core components such as agents, instructions, skills, and plugins, suitable for scenarios like unified team norms and domain-specific development.

With GitHub Copilot's launch of custom agent functionality, the developer community has begun exploring scenario-specific AI assistants. Cogni AI Agentic Collections emerged to expand Copilot's capability boundaries, allowing developers to customize their AI-assisted experience via custom agents and skills based on their needs. The project is maintained by Cogni-AI-OU, sourced from GitHub, and released on May 24, 2026.

The project's core components include:

1. Agents: Specialized AI assistants (e.g., code review, architecture design agents). Installation command: git clone --depth=1 https://github.com/Cogni-AI-OU/cogni-ai-agents ~/.copilot/agents
2. Instructions: Define agent behavior patterns. Installation command: git clone --depth=1 https://github.com/Cogni-AI-OU/cogni-ai-agent-instructions ~/.copilot/instructions
3. Skills: Modular capability units, supporting individual installation (gh skills install Cogni-AI-OU/cogni-ai-agent-skills --scope user <skill-name>) or batch installation
4. Plugins: Advanced encapsulation, supporting installation for Copilot and Claude Code (e.g., Copilot plugin: copilot plugin marketplace add Cogni-AI-OU/cogni-ai-agentic-collections)

Usage Scenarios:

• Unified team coding standards: Ensure consistent code style via instructions and agents
• Domain-specific development: Specialized agents for tech stacks like React, Rust, etc.
• Automated code review: Conduct preliminary reviews as part of CI/CD
• Newcomer training assistance: Answer questions about codebase structure, etc.

Ecosystem Integration: Deeply compatible with the GitHub Copilot ecosystem (custom agent API, Copilot CLI, etc.), while supporting Claude Code, allowing configuration sharing across tools.

Development Environment: Built-in .devcontainer configuration provides a consistent environment (including GitHub Actions, Docker, etc.). Learning Resources: Includes links to GitHub's official custom agent documentation, tutorials, Copilot CLI documentation, etc. License: Uses the MIT License, allowing commercial/non-commercial use, modification, and distribution (with attribution required).

Before using, note the following:

• Learning Curve: Requires understanding the principles of Copilot's agent system
• Maintenance Burden: Custom configurations need to be updated in sync with tool updates
• Team Consistency: Need to establish a unified configuration management strategy
• Feature Dependencies: Some features depend on specific versions of Copilot/Claude Code

Cogni AI Agentic Collections is a practical extension collection in the Copilot ecosystem, lowering the threshold for teams to customize their AI-assisted experience. Whether used directly or as a reference, it provides value to developers. As AI programming tools evolve, such community contributions will become even more important.