Context Compressor — Real-time conversation context compression engine. Activates when user says "compress context", "压缩上下文", "上下文爆了", "context full", "上下文瘦身", "reduce context", or when context exceeds 70% threshold. This skill compresses the CURRENT conversation history into a clean, short prompt. NOT a memory manager — pure context compression tool. Core capabilities: - Compress current chat history (180k → 30k instantly) - Auto-trigger at 70% context threshold - Structured JSON output wit...
Scanned 9/9/2026
Install to Claude Code
npx -y skills add Lord1Egypt/awesome-skill-forge --skill hawk-context --agent claude-codeInstalls into .claude/skills of the current project.
Are you the author of Hawk Context?
Add the live security badge to your README — it updates automatically with every re-scan.
[](https://www.skillsdirectory.com/skills/lord1egypt-hawk-context)More formats (shields.io, HTML) on the badges page.
---
name: context-compressor
description: |
Context Compressor — Real-time conversation context compression engine. Activates when user says "compress context", "压缩上下文", "上下文爆了", "context full", "上下文瘦身", "reduce context", or when context exceeds 70% threshold. This skill compresses the CURRENT conversation history into a clean, short prompt. NOT a memory manager — pure context compression tool.
Core capabilities:
- Compress current chat history (180k → 30k instantly)
- Auto-trigger at 70% context threshold
- Structured JSON output with compression metrics
- Preserve system prompt, keep recent N messages intact
- Summarize older messages with configurable intensity
- No database, no dependencies, pure Python
---
# Context Compressor 🦅
> **纯上下文压缩引擎** — 不是记忆库,是救命工具。
> 当上下文爆炸前,压缩当前对话,继续轻松对话。
---
## 核心定位
| 工具 | 做什么 | 何时用 |
|------|--------|--------|
| **Context Compressor** | 压缩当前对话上下文 | 现在、立刻、当上下文快满 |
| memory-lancedb-pro | 跨会话持久记忆 | 对话之间存取知识 |
| context-hawk | 长期记忆管理 | 日常沉淀 |
**Context Compressor = 急救工具。memory-lancedb-pro/context-hawk = 日常工具。**
---
## 触发方式
**自动触发**(推荐):
- 上下文超过 70% 时自动提示压缩
- 每 10 轮对话自动检查水位
**手动触发**:
- "压缩上下文" / "compress context"
- "上下文爆了" / "context full"
- "瘦身" / "reduce context"
- `/compress`
---
## Quick Start
```bash
# Install skill
openclaw skills install ./context-compressor.skill
# Auto-link command to ~/bin (one-time)
bash ~/.openclaw/workspace/skills/context-compressor/scripts/install.sh
source ~/.bashrc
# Compress current conversation
hawk-compress --level normal --keep 5
# Dry run (preview)
hawk-compress --dry-run --level light --keep 3
# Python API
python3 -c "from hawk_compress import ContextCompressor; \
c = ContextCompressor(keep_recent=5); \
r = c.compress(your_chat_history); \
print(r['stats']['ratio'])"
```
## 压缩原理
### 输入(压缩前)
```
[完整对话历史 — 180k tokens — 爆炸边缘]
System: 你是一个助手...
User: 第一个问题...
Assistant: 第一个回答...
User: 第二个问题...
Assistant: 第二个回答...
... (越来越长)
```
### 输出(压缩后)
```json
{
"compressed_prompt": "...[结构化压缩后的对话]...",
"original_tokens": 180000,
"compressed_tokens": 32000,
"ratio": "5.6x",
"kept_messages": 5,
"summarized_count": 87,
"compression_level": "normal",
"timestamp": "2026-03-29T00:39:00+08:00"
}
```
---
## 压缩层级
| 层级 | 触发 | 效果 | 适用 |
|------|------|------|------|
| `light` | 60-70% | 摘要 > 30天的消息 | 日常维护 |
| `normal` | 70-85% | 摘要 + 保留最近10轮 | 推荐默认 |
| `heavy` | 85-95% | 只保留最近5轮 + 核心摘要 | 紧急急救 |
| `emergency` | > 95% | 只保留系统提示 + 最近3轮 | 立即执行 |
---
## 系统指令永久保留
以下内容**永远不参与压缩**(完整保留):
- System Prompt / SOUL.md / AGENTS.md
- 用户设定的角色定义
- 核心规则和约束
- 当前的任务描述
---
## 重要度过滤
压缩时自动判断每条消息的重要度:
| 重要度 | 消息类型 | 处理方式 |
|--------|---------|---------|
| 🔴 极高 | 决策/规则/任务 | 保留原文 |
| 🟡 高 | 技术方案/代码片段 | 保留摘要 |
| 🟢 中 | 一般讨论 | 摘要或合并 |
| ⚪ 低 | 闲聊/确认/废话 | **直接丢弃** |
---
## 压缩策略
### 1. 消息摘要
把每条老消息压缩为一句话:
```
User: 讨论了Laravel的四层架构,讨论了Controller层的作用...
→ [摘要] User就Laravel四层架构提出问题
```
### 2. 合并重复
重复的说明/确认/指令合并为一条:
```
User: 好的
User: 好的
User: 明白了
→ [合并] User确认理解
```
### 3. 代码折叠
长代码片段只保留文件路径和关键行号:
```
[代码: app/Logic/OrderLogic.php — 45行] → [代码折叠]
```
### 4. 时间戳裁剪
同一时间段内的密集对话压缩为一条:
```
[上午10:00-10:30 共12轮对话] → [摘要]
```
---
## 压缩后的上下文格式
```
## 对话摘要
[最近5轮完整对话]
User: 最新问题...
Assistant: 最新回答...
[历史摘要]
- 2026-03-28: 讨论了Skill架构,决定不拆分
- 2026-03-28: 补充了DAO查询模式
- 2026-03-28: 完成qujin-laravel-team Skill v2
## 任务状态
- 当前任务:压缩上下文
- 进度:进行中
## 核心规则(永久保留)
[系统提示内容]
## 用户偏好(永久保留)
[关键偏好]
```
---
## 参考文档
| 文档 | 用途 |
|------|------|
| [references/compression-logic.md](references/compression-logic.md) | 压缩算法详解 |
| [references/auto-trigger.md](references/auto-trigger.md) | 自动触发机制 |
| [references/structured-output.md](references/structured-output.md) | JSON输出格式 |
| [references/cli.md](references/cli.md) | CLI工具 |
Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.
No comments yet. Be the first to comment!