context-window-management
管理LLM上下文窗口的策略，包括摘要、裁剪、路由以及避免上下文腐化。适用于：上下文窗口、词元上限、上下文管理、上下文工程、长上下文。
机器翻译查看英文原文
Strategies for managing LLM context windows including summarization, trimming, routing, and avoiding context rot Use when: context window, token limit, context management, context engineering, long context.
适用对象：学习与科研。自动收录 · 待核查。
原作者说明节选
从固定版本的 SKILL.md 提取，保留原文语言。以下内容属于作者说明，本站尚未验证安装、兼容或运行结果。
查看英文原文
原文涵盖的主题
- Context Window Management
- Capabilities
- Patterns
- Tiered Context Strategy
- Serial Position Optimization
- Intelligent Summarization
- Anti-Patterns
- ❌ Naive Truncation
You're a context engineering specialist who has optimized LLM applications handling millions of conversations. You've seen systems hit token limits, suffer context rot, and lose critical information mid-dialogue. You understand that context is a finite resource with diminishing returns. More tokens doesn't mean better results—the art is in curating the right information. You know the serial position effect, the lost-in-the-middle problem, and when to summarize versus when to retrieve. Your cor context-engineering context-summarization context-trimming context-routing token-counting context-prioritization Different strategies based on context size Place important content at start and end Summarize by importance, not just recency Works well with: rag-implementation, conversation-memory, prompt-caching, llm-npc-dialogue
- 来源仓库
- davila7/claude-code-templates
- 源文件
- cli-tool/components/skills/ai-research/context-window-management/SKILL.md
- 固定版本
- 21be197f100a12b8d0dd9a211cc244c8179f84d3
来源与内容状态
本条为自动发现的元数据，尚未完成人工文档核查；适用性与运行结果待核查。
原作者：davila7/claude-code-templates · 原作者仓库元数据。文档核查：未核查。最近来源检查：2026-11-09T14:01:43.006Z。
分类由名称、说明与路径规则生成，分类不代表已核实适配。
许可与权利归属
MIT。第三方内容及版权归原作者，使用条件以原文为准。
许可原文 ↗本站没有因同步、复制或用户自报成功而将资源标为运行验证通过。
限制与注意事项
- 元数据收录不代表本站已经安装、适配、实测或进行安全审查。