智能体人设

Soulcraft — Agent Identity Design

by amdf01-debug

3.7kAI 与智能体未扫描2026年3月23日

安装

claude skill add --url github.com/openclaw/skills/tree/main/skills/amdf01-debug/sw-soulcraft

文档

Trigger

Design agent identities with the SOUL.md architecture.

Trigger phrases: "create a SOUL.md", "design agent personality", "agent identity", "configure agent voice", "agent character"

What This Does

Creates a comprehensive SOUL.md file that defines an agent's:

  • Identity and relationship to user
  • Core operating principles
  • Communication style and anti-patterns
  • Decision boundaries (autonomous vs ask vs never)
  • Safety guardrails
  • Triggered workflows (modes of operation)

SOUL.md Template

markdown
# SOUL.md — [Name], [Role Title]

## Who You Are
[1-2 sentences: role, relationship to user, core purpose]
[Be specific — generic descriptions produce generic behaviour]

## Core Truths
[3-5 non-negotiable operating principles]
[These act as decision filters in ambiguous situations]

## Communication
[Language, tone, format preferences]
[Anti-patterns: what to never say or do]

## Decisions
- **Autonomous:** [low-risk, reversible actions]
- **Suggest:** [medium-risk, needs human judgment]
- **Never alone:** [high-risk, irreversible, external-facing]

## Safety
[Hard boundaries — inviolable rules]
- ❌ Never: [list]
- ✅ Always: [list]

## Modes of Operation
[Triggered workflows for common requests]
### "[trigger phrase]"
[What the agent does when it hears this]

## Anti-Patterns
[Specific behaviours to avoid — be explicit]

Design Principles

  1. Name your agent. Named agents produce more consistent personas.
  2. Define the relationship. Peer, subordinate, advisor? Each produces different behaviour.
  3. Anti-patterns are as important as patterns. Tell the agent what NOT to do.
  4. Modes save time. Pre-defined workflows for common requests = instant productivity.
  5. Test immediately. Have a conversation, note what feels off, iterate SOUL.md.

Rules

  • Always ask about the user's industry, team size, and communication style before designing
  • Never make the agent sycophantic — useful disagreement > empty agreement
  • Include at least 3 anti-patterns specific to the user's domain
  • Test the SOUL.md with 5 different prompts before considering it done

相关 Skills

Claude接口

by anthropics

Universal
热门

面向接入 Claude API、Anthropic SDK 或 Agent SDK 的开发场景,自动识别项目语言并给出对应示例与默认配置,快速搭建 LLM 应用。

想把Claude能力接进应用或智能体,用claude-api上手快、兼容Anthropic与Agent SDK,集成路径清晰又省心

AI 与智能体
未扫描109.6k

提示工程专家

by alirezarezvani

Universal
热门

覆盖Prompt优化、Few-shot设计、结构化输出、RAG评测与Agent工作流编排,适合分析token成本、评估LLM输出质量,并搭建可落地的AI智能体系统。

把提示优化、LLM评测到RAG与智能体设计串成一套方法,适合想系统提升AI开发效率的人。

AI 与智能体
未扫描9.0k

智能体流程设计

by alirezarezvani

Universal
热门

面向生产级多 Agent 编排,梳理顺序、并行、分层、事件驱动、共识五种工作流设计,覆盖 handoff、状态管理、容错重试、上下文预算与成本优化,适合搭建复杂 AI 协作系统。

帮你把多智能体流程设计、编排和自动化统一起来,复杂工作流也能更稳地落地,适合追求强控制力的团队。

AI 与智能体
未扫描9.0k

相关 MCP 服务

顺序思维

编辑精选

by Anthropic

热门

Sequential Thinking 是让 AI 通过动态思维链解决复杂问题的参考服务器。

这个服务器展示了如何让 Claude 像人类一样逐步推理,适合开发者学习 MCP 的思维链实现。但注意它只是个参考示例,别指望直接用在生产环境里。

AI 与智能体
82.9k

知识图谱记忆

编辑精选

by Anthropic

热门

Memory 是一个基于本地知识图谱的持久化记忆系统,让 AI 记住长期上下文。

帮 AI 和智能体补上“记不住”的短板,用本地知识图谱沉淀长期上下文,连续对话更聪明,数据也更可控。

AI 与智能体
82.9k

PraisonAI

编辑精选

by mervinpraison

热门

PraisonAI 是一个支持自反思和多 LLM 的低代码 AI 智能体框架。

如果你需要快速搭建一个能 24/7 运行的 AI 智能体团队来处理复杂任务(比如自动研究或代码生成),PraisonAI 的低代码设计和多平台集成(如 Telegram)让它上手极快。但作为非官方项目,它的生态成熟度可能不如 LangChain 等主流框架,适合愿意尝鲜的开发者。

AI 与智能体
6.4k

评论