How does this skill help with multi-agent systems?

It provides established architectural patterns like supervisor/orchestrator and peer-to-peer swarms to help isolate context and improve task decomposition.

Is this skill compatible with frameworks like LangGraph or CrewAI?

Yes, while designed for Claude Code, the architectural patterns and optimization principles are platform-agnostic and apply to any production agent framework.

Can this skill reduce LLM token costs?

Yes, it includes strategies for context compression, compaction, and strategic partitioning designed to lower the 'tokens-per-task' ratio.

What is context engineering in AI agents?

Context engineering is the process of curating the complete state available to a language model—including history, tool outputs, and retrieved data—to maximize relevant signals and minimize noise.

Context Engineering for AI Agents

Name: Context Engineering for AI Agents
Author: ken-cavanagh-glean

byken-cavanagh-glean

•

Data Science & ML

Provides structured patterns and guidance for building high-performance, production-grade AI agent systems through advanced context management.

About

This skill collection equips developers with the foundational and architectural frameworks necessary to manage LLM context windows effectively, preventing information degradation in complex AI applications. It covers critical patterns for multi-agent coordination—such as supervisor and swarm architectures—alongside advanced memory system designs ranging from vector RAG to knowledge graphs. By implementing these strategies, developers can optimize token usage through context compression and masking, ensuring that agents maintain high signal-to-noise ratios even during long-running sessions or complex task decompositions.

Key Features

Context optimization techniques such as compression, compaction, and observation masking.
Comprehensive evaluation frameworks for measuring agent accuracy, efficiency, and process quality.
2 GitHub stars
Multi-agent coordination patterns including supervisor, swarm, and hierarchical architectures.
Strategic tool design principles for robust deterministic and non-deterministic interfaces.
Advanced memory system design utilizing vector RAG, knowledge graphs, and file-system-as-memory.

Use Cases

Debugging context-related failures and 'lost-in-middle' phenomena in LLM applications.
Optimizing LLM token efficiency and performance in production-scale agent sessions.
Designing and implementing multi-agent systems for complex software engineering tasks.

About

Key Features

Context optimization techniques such as compression, compaction, and observation masking.
Comprehensive evaluation frameworks for measuring agent accuracy, efficiency, and process quality.
2 GitHub stars
Multi-agent coordination patterns including supervisor, swarm, and hierarchical architectures.
Strategic tool design principles for robust deterministic and non-deterministic interfaces.
Advanced memory system design utilizing vector RAG, knowledge graphs, and file-system-as-memory.

Use Cases

Debugging context-related failures and 'lost-in-middle' phenomena in LLM applications.
Optimizing LLM token efficiency and performance in production-scale agent sessions.
Designing and implementing multi-agent systems for complex software engineering tasks.