context-degradation

Diagnoses context degradation issues in LLMs and suggests mitigation strategies.

Updated Aug 27, 2026
One-click install
npx skills add https://github.com/vaur94/opencode-multiagent --skill context-degradation-vaur94
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: context-degradation
Source: https://github.com/vaur94/opencode-multiagent/tree/main/skills/context-degradation
Command: npx skills add https://github.com/vaur94/opencode-multiagent --skill context-degradation-vaur94

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve?

This Skill helps diagnose and mitigate issues arising from context degradation in large language models, such as the "lost in middle" phenomenon, context poisoning, and confusion, which lead to agent performance degradation.

Core Features & Use Cases

  • Diagnose Context Failures: Identifies patterns like lost-in-middle, poisoning, distraction, confusion, and clash.
  • Mitigation Strategies: Provides architectural patterns and practical guidance for managing context effectively.
  • Use Case: An agent is consistently failing on tasks that involve long documents. Activating this skill helps understand if the failure is due to context degradation and suggests solutions like summarizing or partitioning context.

Quick Start

Use the context-degradation skill to understand why my agent is failing on long documents.

Frequently Asked Questions about context-degradation

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
Why does my LLM agent fail when processing long documents?▼

Context degradation occurs when lost-in-middle phenomena, context poisoning, or attention confusion degrade LLM agent performance on long documents. This Skill diagnoses these specific attention pattern failures and token budget limitations.

How do I fix the lost in middle context issue in large language models?▼

Fix the lost in middle issue by applying architectural patterns for effective context management, such as summarizing or partitioning context. These mitigation strategies prevent attention pattern failures within limited token budgets.

What is context poisoning and how does it affect agent performance?▼

Context poisoning is a degradation pattern where corrupted information disrupts LLM attention mechanisms, leading to agent confusion and performance degradation. Diagnosing this attention pattern helps determine if context partitioning is needed.

How do I diagnose context degradation issues in my LLM?▼

Diagnose context degradation by analyzing agent failures for patterns like lost-in-middle, poisoning, distraction, confusion, and clash. Understanding attention mechanisms and token budget limitations is required to identify these specific failure modes.

What are the best architectural patterns for managing LLM context effectively?▼

Effective context management architectural patterns include summarizing and partitioning context to prevent degradation. Implementing these strategies requires understanding attention mechanisms and token budget limitations to maintain agent performance.

When should I partition context to prevent LLM agent failures?▼

Partition context to prevent LLM agent failures when diagnosing context degradation issues like lost-in-middle or context poisoning. This mitigation strategy effectively manages token budget limitations and restores attention pattern reliability.