Is this skill specifically for DeepSeek models?

While it is specifically tuned for DeepSeek's 64k context window, its parameters are fully configurable to support any LLM context limit.

How does Context Optimizer prevent context overflow?

It uses four distinct compaction strategies—semantic, temporal, extractive, and adaptive—to merge, summarize, or prune messages based on their importance and relevance to the current task.

What is Query-aware Relevance scoring?

This feature analyzes your current prompt and scores previous messages based on their similarity to your request, ensuring the most relevant context is prioritized for the AI's current focus.

Does this skill delete my conversation history?

No, it uses a hierarchical memory system. Information pruned from the 'active' context is moved to a searchable archive, allowing it to be retrieved later if needed.

Context Optimizer

Name: Context Optimizer
Author: sundial-org

bysundial-org

•

150

•

生产力与工作流

Optimizes AI context windows through intelligent pruning, semantic compaction, and hierarchical memory management for long-running sessions.

The Context Optimizer is a sophisticated management tool specifically designed to maximize the efficiency of large context windows, such as DeepSeek's 64k limit, by preventing token overflow while preserving critical information. It utilizes a multi-strategy approach including semantic merging, temporal summarization, and extractive compaction to reduce message volume without losing core meaning. With its unique hierarchical memory system, it functions like a RAM-and-storage architecture, keeping relevant data in the active window while archiving older conversations into a searchable long-term storage for later retrieval. This ensures that Claude remains focused, responsive, and informed during even the most complex development tasks.

主要功能

01Hierarchical memory system with a searchable 100MB context archive

02Real-time health monitoring with automated logging of optimization events to chat

03Query-aware dynamic relevance scoring to prioritize high-value information

04DeepSeek-optimized 64k context window management and token tracking

05Multi-strategy auto-compaction including semantic, temporal, and extractive methods

06150 GitHub stars

使用场景

01Managing extremely long coding sessions without hitting context window limits

02Reducing API costs and improving response speed by pruning redundant message data

03Retrieving specific historical project details from a long-term conversation archive

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add sundial-org/awesome-openclaw-skills context-optimizer

For use in Claude.ai and ChatGPT

Download Skill