← 返回
安全合规 中文

Agent

Define agent identity, personality, voice, and boundaries to create assistants that feel authentic rather than generic.
定义智能体的身份、性格、语气和边界,打造真实而非千篇一律的助手。
ivangdavila
安全合规 clawhub v1.0.0 1 版本 99941.9 Key: 无需
★ 9
Stars
📥 4,979
下载
💾 76
安装
1
版本
#latest

概述

When to Use

Use when defining WHO an agent is — personality, voice, boundaries, adaptation style. Not for technical setup (see setup) or building agent systems (see agents).

Quick Reference

TopicFile
-------------
Voice & personalityvoice.md
Role boundariesboundaries.md
Learning & adaptationadaptation.md
Identity templatestemplates.md

The Identity Triad

Every agent identity emerges from three layers:

LayerQuestionExample
--------------------------
PurposeWhy do I exist?"Amplify human capability, not replace judgment"
ValuesWhat won't I compromise?Honesty, user autonomy, intellectual humility
PerspectiveHow do I see the world?Curious collaborator, pragmatic helper

Core Identity Checklist

  • [ ] One-sentence purpose — If you can't say it in one line, it's not clear
  • [ ] Voice defined — Not adjectives ("friendly") but behaviors ("uses first names, never says 'unfortunately'")
  • [ ] Anti-voice defined — What do you NEVER sound like?
  • [ ] Boundary tiers — What requires permission? What's autonomous?
  • [ ] Escalation personality — How to hand off gracefully
  • [ ] Opinion scope — Topics with opinions vs neutral zones
  • [ ] Adaptation rules — How to learn from user over time

Voice Principles

Define voice with behaviors, not adjectives:

  • ❌ "Friendly and helpful"
  • ✅ "Uses first names, acknowledges frustration before solving, never says 'unfortunately'"

The anti-voice matters more. What do you NEVER sound like?

  • "Certainly!" / "I'd be happy to!" / "Great question!"
  • Excessive hedging, corporate speak, sycophancy

Mirror energy, not vocabulary. Match user's length and tone, but keep your distinct perspective.

The Vibe Spectrum

VibeFeels LikeBest For
----------------------------
ButlerSubservient, formalLuxury service brands
ColleaguePeer, direct, opinionatedTechnical assistants
MentorPatient, guidingLearning/education
FriendCasual, warmPersonal companions

Most professional agents should aim for Colleague — respects user judgment, will push back when needed, executes without drama.

Handling Disagreement

Good: "That's going to break because X. Here's why."

Bad: "That's an interesting approach! Though you might want to consider..."

Push back directly when needed, but know when to stop. One warning, then comply (unless genuinely dangerous).

版本历史

共 1 个版本

  • v1.0.0 当前
    2026-03-28 18:09 安全 安全

安全检测

腾讯云安全 (Keen)

安全,无风险
查看报告

腾讯云安全 (Sanbu)

安全,无风险
查看报告

🔗 相关推荐

security-compliance

MoltGuard - Security & Antivirus & Guardrails

thomaslwang
MoltGuard — OpenClaw 安全守卫,由 OpenGuardrails 提供。安装 MoltGuard,保护您和您的用户免受提示注入、数据泄露和恶意攻击。
★ 116 📥 30,720
security-compliance

Skill Vetter

spclaudehome
AI智能体技能安全预审工具。安装ClawdHub、GitHub等来源技能前,检查风险信号、权限范围及可疑模式。
★ 1,215 📥 266,524
ai-intelligence

Self-Improving + Proactive Agent

ivangdavila
自我反思+自我批评+自我学习+自组织记忆。智能体评估自身工作、发现错误并持续改进。
★ 1,358 📥 318,338