Can this skill help with context-related debugging?

Yes, it provides specific metrics and identifies anti-patterns like Snapshot Broadcasting or Defensive Inclusion that cause agent performance degradation.

What is the primary benefit of Context Engineering?

The primary benefit is a 60-80% reduction in token consumption, which lowers costs and improves agent reasoning speed by removing irrelevant data.

How does tiered fidelity work?

It assigns different levels of data detail (Full, Selective, Filtered, Minimal, or Metadata) to specific agents based on their functional requirements.

What are the Four Laws of Context Management?

They are Selective Projection, Tiered Fidelity, Reference vs. Embedding, and Lazy Loading, which together ensure agents only process the minimum necessary data.

What is a 'NOT PASSED' section in context management?

It is an explicit declaration of what data has been excluded during an agent handoff to prevent the agent from hallucinating or searching for missing context.

Context Engineering

Name: Context Engineering
Author: abossenbroek

byabossenbroek

开发者工具

Optimizes LLM agent performance and reduces token consumption through advanced context management and tiered fidelity patterns.

关于

The Context Engineering skill provides a comprehensive framework for implementing state-of-the-art context management within multi-agent workflows. By applying the 'Four Laws of Context Management'—Selective Projection, Tiered Fidelity, Reference vs. Embedding, and Lazy Loading—it enables developers to achieve 60-80% token reduction. This skill is essential for debugging context-related issues, creating efficient agent handoff protocols, and maintaining high reasoning performance in complex, data-heavy AI systems by ensuring each agent receives only the information necessary for its specific role.

主要功能

0 GitHub stars
Standardizes agent handoff protocols using structured, high-efficiency YAML schemas.
Implements the 'Four Laws' of context management to minimize token overhead and waste.
Provides severity-based batching to optimize and reduce validation operations.
Defines standardized context tiers ranging from Full to Metadata for role-specific data delivery.
Enforces 'NOT PASSED' isolation patterns to prevent context leakage and agent confusion.

使用场景

Architecting efficient communication and data passing between specialized AI agents in a pipeline.
Debugging context-length limitations or 'lost in the middle' issues in complex reasoning tasks.
Reducing token costs and latency in large-scale multi-agent software development workflows.

关于

主要功能

0 GitHub stars
Standardizes agent handoff protocols using structured, high-efficiency YAML schemas.
Implements the 'Four Laws' of context management to minimize token overhead and waste.
Provides severity-based batching to optimize and reduce validation operations.
Defines standardized context tiers ranging from Full to Metadata for role-specific data delivery.
Enforces 'NOT PASSED' isolation patterns to prevent context leakage and agent confusion.

使用场景

Architecting efficient communication and data passing between specialized AI agents in a pipeline.
Debugging context-length limitations or 'lost in the middle' issues in complex reasoning tasks.
Reducing token costs and latency in large-scale multi-agent software development workflows.