RUNE — Prompt Amplification Framework
RUNE transforms flat, ambiguous prompts into structured XML prompts validated by Spinoza's philosophical framework — resulting in outputs that are ~45% higher quality than raw prompting.
The 8 Layers
| Layer | Name | Purpose |
|---|
| L0 | System Core | Role, persona, behavioral rules |
| L1 |
Context Identity | Domain knowledge, target audience |
| L2 | Intent Scope | Task definition, output format |
| L3 | Governance | Constraints, ethical boundaries |
| L4 | Cognitive Engine | Thinking strategy (CoT, ToT) |
| L5 | Capabilities Domain | Tools, integrations, capabilities |
| L6 | QA | Validation criteria, quality control |
| L7 | Output Meta | Format, style, length, language |
Requirements
- - Python 3.11+
- RUNE repo cloned locally
- INLINECODE0 in INLINECODE1
Usage
CODEBLOCK0
Setup
CODEBLOCK1
Source
- - Author: NeuraByte Labs
- Version: RUNE v4.3 / WAND v1.5.0
- Repo: https://github.com/neurabytelabs/rune-skill
RUNE — 提示词放大框架
RUNE 能将扁平、模糊的提示词转换为经过斯宾诺莎哲学框架验证的结构化XML提示词——与原始提示相比,输出质量提升约45%。
8层结构
上下文身份 | 领域知识、目标受众 |
| L2 | 意图范围 | 任务定义、输出格式 |
| L3 | 治理层 | 约束条件、伦理边界 |
| L4 | 认知引擎 | 思维策略(CoT、ToT) |
| L5 | 能力领域 | 工具、集成、能力 |
| L6 | 质量保障 | 验证标准、质量控制 |
| L7 | 输出元数据 | 格式、风格、长度、语言 |
环境要求
- - Python 3.11+
- 本地克隆RUNE仓库
- ~/.secrets文件中配置RUNEAPIKEY
使用方法
bash
管道输入提示词
echo 写一篇关于AI的博客文章 | bash main.sh
作为参数传入
bash main.sh 用量子纠缠向12岁孩子解释
安装步骤
bash
1. 克隆RUNE仓库
git clone https://github.com/mrsarac/master-prompts ~/Documents/GitHub/rune
2. 将API密钥添加到~/.secrets
echo export RUNE
APIKEY=你的密钥 >> ~/.secrets
3. 测试
echo 你好 | bash main.sh
来源
- - 作者: NeuraByte Labs
- 版本: RUNE v4.3 / WAND v1.5.0
- 仓库: https://github.com/neurabytelabs/rune-skill