KIMGCA Agent 配置方法论文档
KIMGCA 是一种 Agent 配置方法论,融合了:
- 知识继承 — 从现有知识库继承 Gene/Capsule
- 模块化架构 — Gene/Capsule 可独立加载、组合、扩展
- 角色驱动设计 — 基于明确的目标和职责构建 Agent
- 分层引导配置 — Identity→SOUL→AGENTS→MEMORY→USER 五层结构
KIMGCA is an Agent configuration methodology that combines:
- Knowledge Inheritance — Inherit Gene/Capsule from existing knowledge bases
- Modular Architecture — Gene/Capsule can be independently loaded, combined, and extended
- Role-Driven Design — Build Agent based on clear goals and responsibilities
- Layered Bootstrap Configuration — Five-layer structure: Identity→SOUL→AGENTS→MEMORY→USER
KIMGCA 源自 EvoMap.ai 平台的学习与实践。
We developed KIMGCA through learning and practice on the EvoMap.ai platform.
我们的学习历程:
- 发现 EvoMap.ai 平台,学习其 Gene/Capsule 架构
- 在 OpenClaw 框架上实现 EvoMap 概念
- 结合 msitarzewski 的基因理论进行扩展
- 形成 KIMGCA 方法论
Our Learning Journey:
- Discovered EvoMap.ai platform, studied its Gene/Capsule architecture
- Implemented EvoMap concepts on OpenClaw framework
- Extended using msitarzewski's Gene theory
- Formed KIMGCA methodology
免责声明: 我们仍在学习 EvoMap.ai 平台的精髓,尚未完全掌握。
Disclaimer: We are still learning the essence of EvoMap.ai platform and have not fully mastered it.
| 来源 | 描述 | 角色 |
|---|---|---|
| EvoMap.ai | Gene/Capsule 架构发源地,核心概念来源 | 主框架标准 |
| OpenClaw | Agent 运行框架,技术实现基础 | 技术框架 |
| msitarzewski | AI Agent 设计专家,98K+ stars,提供 Gene/Capsule 内容 | 内容来源 |
| RedAgentTeam | KIMGCA 框架原创贡献 | 框架整合 |
| Source | Description | Role |
|---|---|---|
| EvoMap.ai | Origin of Gene/Capsule architecture, core concept source | Primary framework |
| OpenClaw | Agent runtime framework, technical foundation | Technical framework |
| msitarzewski | AI Agent design expert, 98K+ stars, provides Gene/Capsule content | Content source |
| RedAgentTeam | KIMGCA original contribution | Framework integration |
kimgca/
├── README.md # 本文件
├── docs/ # 文档目录
│ └── 01-methodology.md # 方法论详细文档
├── templates/ # 配置模板
│ ├── 01-identity-template.md
│ ├── 02-soul-template.md
│ ├── 03-agents-template.md
│ ├── 04-memory-template.md
│ └── 05-user-template.md
└── examples/ # 配置示例
└── evoagent-example.md # evoagent 实际配置示例
| 层级 | 文件 | 职责 |
|---|---|---|
| Identity | IDENTITY.md | 身份定义 |
| Soul | SOUL.md | 核心信念 |
| Agents | AGENTS.md | 行为准则 |
| Memory | MEMORY.md | 记忆架构 |
| User | USER.md | 用户画像 |
Gene 是可复用的策略模板,包含触发条件、行为准则、执行流程、验证标准。
Gene is a reusable strategy template containing trigger conditions, behavior guidelines, execution flow, and verification standards.
Capsule 是验证过的执行路径,包含输入输出定义、执行步骤、错误处理、成功标准。
Capsule is a validated execution path containing input/output definitions, execution steps, error handling, and success criteria.
mkdir -p ~/.openclaw/workspace/<agent-name>/
cd ~/.openclaw/workspace/<agent-name>/
mkdir -p genes capsules memory logs复制 templates/ 中的模板到目标目录。
Copy templates from templates/ to your agent directory.
从 RedAgentTeamWiki/msitarzewski/genes/ 选择需要的 Gene。
Select needed Genes from RedAgentTeamWiki/msitarzewski/genes/.
{
"id": "<agent-id>",
"workspace": "~/.openclaw/workspace/<agent-name>",
"model": {
"primary": "evomap/evomap-kimi-k2.6",
"fallbacks": ["minimax/MiniMax-M2.7", "freemodel/gpt-5.4"]
}
}| 文档 | 说明 |
|---|---|
| docs/01-methodology.md | KIMGCA 方法论详细文档(v1.1.0) |
| templates/*.md | 五层配置模板 |
| examples/evoagent-example.md | evoagent 实际配置示例 |
| 版本 | 日期 | 说明 |
|---|---|---|
| 1.0.0 | 2026-05-26 | 首次文档化 |
| 1.0.1 | 2026-05-26 | 增加「局限性」和「术语来源」章节 |
| 1.1.0 | 2026-05-26 | 增加「Sovereign Evolution」和「防幻觉机制」章节 |
| 1.2.0 | 2026-05-26 | 中英双语 + EvoMap.ai 主推 + 学习现状 + LICENSE |
- evoagent — 基于 KIMGCA 配置的 Agent 示例
- EvoMap.ai — Gene/Capsule 架构发源地
- OpenClaw Wiki — OpenClaw 官方文档
我们正在努力学习和掌握 EvoMap.ai 平台的真谛。
We are striving to learn and master the essence of the EvoMap.ai platform.
当前状态:
- ✅ 理解基本概念(GEP、Gene、Capsule、Evolver、Credits)
⚠️ 尚未完全掌握高级特性⚠️ 实践中仍在探索最佳实践
Current Status:
- ✅ Understanding basic concepts (GEP, Gene, Capsule, Evolver, Credits)
⚠️ Not yet fully mastering advanced features⚠️ Still exploring best practices in implementation
我们诚实地说明:
- KIMGCA 框架是我们在 EvoMap.ai 启发下的原创整合
- Gene/Capsule 内容来自 msitarzewski(已归因)
- 如有错误或不足,欢迎指正
We honestly state:
- KIMGCA framework is our original integration inspired by EvoMap.ai
- Gene/Capsule content comes from msitarzewski (attributed)
- We welcome corrections and suggestions
文档状态: 草稿
维护者: RedAgentTeam
最后更新: 2026-05-26