Optimize LLM performance with 4 distinct context tiers.
Learn summarization and routing to avoid context rot and reduce token costs. Download now.
LLM Memory
Tokenization
Prompt Engineering
Claude API
Context Engineering
About This Skill
This skill provides 6 core capabilities for managing LLM memory, including 4 tiered strategies ranging from 8,000 to over 100,000 tokens. Master context engineering to prevent rot and optimize performance.
Quick Start
1Install tiktoken for accurate token counting
2Define your ContextTier thresholds for different models
3Implement the selectStrategy function to route messages
Example Command
npx antigravity-cli add context-window-management
Core Capabilities
Context Trimming
Remove redundant or low-priority tokens to fit within model limits while maintaining semantic meaning.
Context Summarization
Condense long conversation histories into concise summaries to fit within large context windows.
Context Routing
Direct prompts to different models based on current token usage and conversation complexity.
Token Counting
Accurately measure context size using tools like tiktoken for precise management and cost control.
Usage Examples
Before
35,000 tokens of raw chat history causing 'context rot' and high latency.
After
7,500 tokens of summarized context with 100% factual retention.
Input
User asks a simple question with a large history attached.
Output
System routes the request to Claude-3-Haiku using the 8k 'full' strategy.
Input
System detects context approaching 100k tokens.
Output
Triggers the 'summarize' strategy to condense history before the next turn.
SKILL.md
---
name: context-window-management
description: Strategies for managing LLM context windows including
summarization, trimming, routing, and avoiding context rot
risk: unknown
source: vibeship-spawner-skills (Apache 2.0)
date_added: 2026-02-27
---
# Context Window Management
Strategies for managing LLM context windows including summarization, trimming, routing, and avoiding context rot
## Capabilities
- context-engineering
- context-summarization
- context-trimming
- context-routing
- token-counting
- context-prioritization
## Prerequisites
- Knowledge: LLM fundamentals, Tokenization basics, Prompt engineering
- Skills_recommended: prompt-engineering
## Scope
- Does_not_cover: RAG implementation details, Model fine-tuning, Embedding models
- Boundaries: Focus is context optimization, Covers strategies not specific implementations
## Ecosystem
### Primary_tools
- tiktoken - OpenAI's tokenizer for counting tokens
- LangChain - Framework with context management utilities
- Claude API - 200K+ context with caching support
## Patterns
### Tiered Context Strategy
Different strategies based on context size
**When to use**: Building any multi-turn conversation system
interface ContextTier {
maxTokens: number;
strategy: 'full' | 'summarize' | 'rag';
model: string;
}
const TIERS: ContextTier[] = [
{ maxTokens: 8000, strategy: 'full', model: 'claude-3-haiku' },
{ maxTokens: 32000, strategy: 'full', model: 'claude-3-5-sonnet' },
{ maxTokens: 100000, strategy: 'summarize', model: 'claude-3-5-sonnet' },
{ maxTokens: Infinity, strategy: 'rag', model: 'claude-3-5-sonnet' }
];
async function selectStrategy(messages: Message[]): ContextTier {
const tokens = await countTokens(messages);
for (const tier of TIERS) {
if (tokens <= tier.maxTokens) {
return tier;
}
}
return TIERS[TIERS.length - 1];
}
async function prepareContext(messages: Message[]): PreparedContext {
const tier = await select
Frequently Asked Questions
FAQ
Which tools are compatible with this skill?
This skill is designed to work with tiktoken for tokenization, LangChain for orchestration, and the Claude API for high-capacity context handling.
Who is the target audience for this skill?
Software developers and AI engineers building multi-turn conversational agents or complex RAG systems.
How does this differ from standard RAG?
While RAG focuses on external retrieval, this skill focuses on managing the active memory (context window) of the LLM itself to prevent degradation.
Does this support multiple programming languages?
The patterns are language-agnostic, though examples are provided in TypeScript/JavaScript for use with Node.js environments.
What results can I expect after implementation?
Expect reduced token costs, lower latency, and significantly reduced 'context rot' where the model forgets early instructions.
Discussion
Discussion
0 comments
U
On This Page
Trigger Phrases
Use these phrases to activate this skill in your AI coding assistant:
