Two free online books explaining how agent harnesses for code-writing models, such as Claude Code and Codex, are designed and controlled.
muratcankoylan/Agent-Skills-for-Context-Engineering
A collection of Agent Skills teaching context engineering, multi-agent design, and evaluation for building production AI agent systems.
- Stars
- 18,315
- Forks
- 1,510
- Open issues
- 57
- Last commit
- 26 days ago
What is it?
This project is a set of Agent Skills, each a SKILL.md-based instruction package, covering context engineering for LLM agents. Topics include context fundamentals, degradation patterns such as lost-in-the-middle, compression, and optimization, along with multi-agent patterns, memory systems, tool design, filesystem-based context, and evaluation including LLM-as-a-judge.
The skills are written to be platform-agnostic and are organized in a repo-root skills/ directory. The repository ships as a Claude Code plugin marketplace and as an Open Plugins package for hosts such as Cursor and Codex, and individual skill folders can be copied into a project's skills directory.
The repo also includes worked examples such as an LLM-as-judge TypeScript implementation, a book style fine-tuning pipeline, and a personal knowledge system, plus a researcher directory with tooling for maintaining the skill corpus.
Key features
- 01Covers context fundamentals, degradation, compression, and optimization for LLM agents.
- 02Includes skills for multi-agent orchestration, memory systems, and tool design.
- 03Provides evaluation skills such as LLM-as-a-judge, pairwise comparison, and rubric generation.
- 04Installable into Claude Code, Cursor, Codex, and other Agent Skills hosts.
- 05Ships worked examples with Python and TypeScript code.
Use cases
- Giving a coding agent reusable guidance on managing long conversations and tool outputs.
- Designing a multi-agent system and deciding how to split context between agents.
- Building an evaluation pipeline with LLM judges for agent outputs.
Quick start
- 1
Register this repository as a plugin marketplace in Claude Code
/plugin marketplace add muratcankoylan/Agent-Skills-for-Context-Engineering - 2
Install the context-engineering plugin
/plugin install context-engineering@context-engineering-marketplace
In collections
Topics
FAQ
- What is Agent-Skills-for-Context-Engineering?
- A collection of Agent Skills teaching context engineering, multi-agent design, and evaluation for building production AI agent systems. This project is a set of Agent Skills, each a SKILL.md-based instruction package, covering context engineering for LLM agents. Topics include context fundamentals, degradation patterns such as lost-in-the-middle, compression, and optimization, along with multi-agent patterns, memory systems, tool design, filesystem-based context, and evaluation including LLM-as-a-judge.
- Is Agent-Skills-for-Context-Engineering free to use? What license does it have?
- Agent-Skills-for-Context-Engineering is open source under the MIT license. Check the license terms before using it in commercial products.
- Is Agent-Skills-for-Context-Engineering still maintained?
- Actively maintained. The last commit to Agent-Skills-for-Context-Engineering was on Oct 12, 2026.
- What can I use Agent-Skills-for-Context-Engineering for?
- Giving a coding agent reusable guidance on managing long conversations and tool outputs. Designing a multi-agent system and deciding how to split context between agents. Building an evaluation pipeline with LLM judges for agent outputs.
- How do I install Agent-Skills-for-Context-Engineering?
- Register this repository as a plugin marketplace in Claude Code: /plugin marketplace add muratcankoylan/Agent-Skills-for-Context-Engineering Install the context-engineering plugin: /plugin install context-engineering@context-engineering-marketplace
- Where can I get Agent-Skills-for-Context-Engineering?
- The source code and installation instructions are on GitHub: https://github.com/muratcankoylan/Agent-Skills-for-Context-Engineering
Summary generated by AI from the project's README. GitHub stats refresh daily.
Related projects
An open-source context service that keeps decisions, constraints, and work state outside chat sessions so agents and people can resume work across sessions.
A dependency manager for AI coding agent configuration, letting teams declare skills, prompts, agents, plugins and MCP servers in one apm.yml file.
A Claude Code plugin that designs domain-specific agent teams and generates the agent definitions and skills they use.
A collection of 15 squads with 183 specialized AI agents, plus workflows, tasks and checklists, distributed as Agent Skills for coding assistants.
A self-learning context layer that gives coding agents like Claude Code and Cursor a map of your codebase plus the decisions your team has made about it.