A systematic skill library for AI Agent context management, covering fundamentals, architectural patterns, operational optimization, and evaluation systems. Compatible with Claude Code, Cursor, and other platforms. Distinguished from prompt engineering by its holistic approach to curating all information entering the model's attention budget.
Agent Skills for Context Engineering is a comprehensive, open collection of skills focused on context engineering principles. Context Engineering is the discipline of managing the language model's context window, addressing the holistic curation of all information entering the model's limited attention budget: system prompts, tool definitions, retrieved documents, message history, and tool outputs.
Foundational Skills
| Skill | Description |
|---|
context-fundamentals | Understand what context is, why it matters, and the anatomy of context in agent systems |
context-degradation | Recognize patterns of context failure: lost-in-middle, poisoning, distraction, and clash |
context-compression | Design and evaluate compression strategies for long-running sessions |
Architectural Skills
| Skill | Description |
|---|
multi-agent-patterns | Master orchestrator, peer-to-peer, and hierarchical multi-agent architectures |
memory-systems | Design short-term, long-term, and graph-based memory architectures |
tool-design | Build tools that agents can use effectively |
filesystem-context | Use filesystems for dynamic context discovery, tool output offloading, and plan persistence |
hosted-agents | Build background coding agents with sandboxed VMs, pre-built images, multiplayer support |
Operational Skills
| Skill | Description |
|---|
context-optimization | Apply compaction, masking, and caching strategies |
evaluation | Build evaluation frameworks for agent systems |
advanced-evaluation | Master LLM-as-a-Judge techniques: direct scoring, pairwise comparison, rubric generation |
Development & Cognitive Skills
project-development: Design and build LLM projects from ideation through deploymentbdi-mental-states: Transform external RDF context into agent mental states (beliefs, desires, intentions)
- Progressive Disclosure: At startup, agents load only skill names and descriptions; full content loads only when activated, saving tokens
- Platform Agnosticism: Based on concepts and Python pseudocode, adaptable to Claude Code, Cursor, Codex, and other platforms
- Conceptual Foundation with Practical Examples: Python pseudocode that works across environments
| Component | Description |
|---|
| System Prompts | Agent's core identity, constraints, behavioral guidelines; loaded once at session start |
| Tool Definitions | Actions an agent can take; includes name, description, parameters, return format |
| Retrieved Documents | Domain-specific knowledge via RAG; just-in-time loading pattern |
| Message History | Conversation state; can dominate context usage in long-running tasks |
| Tool Outputs | Results of agent actions; can reach 83.9% of total context usage |
- Attention Budget Constraint: For n tokens, creates n² relationships
- "Lost-in-the-Middle" Phenomenon: Center of context receives less attention
- U-Shaped Attention Curves: Prioritizes beginning and end
- digital-brain-skill: Personal operating system for founders and creators with 6 modules, 4 automation scripts
- x-to-book-system: Multi-agent system monitoring X accounts and generating daily synthesized books
- llm-as-judge-skills: Production-ready LLM evaluation tools with TypeScript implementation, 19 passing tests
- book-sft-pipeline: Train models to write in any author's style
# Claude Code Plugin Marketplace
/plugin marketplace add muratcankoylan/Agent-Skills-for-Context-Engineering
/plugin install context-engineering-fundamentals@context-engineering-marketplace
/plugin install agent-architecture@context-engineering-marketplace
/plugin install agent-evaluation@context-engineering-marketplace
For Cursor/IDE users, copy skill content into .rules files or create project-specific Skills folders.