Token导航 LogoToken导航

Agent Skill · 开发

claudebuddy(claudebuddy 开发)

claudebuddy 用于记录任务执行中的错误、用户纠正、经验和能力缺口,适合在 OpenClaw 中希望让 Agent 持续沉淀问题、修正和最佳实践时使用。可结合来源仓库、安装命令和原始 README 继续核验具体用法。安装前建议确认权限范围、维护状态,以及是否会触发联网、命令执行或文件读写。

claudebuddy 是开发类 Agent Skill,可用宿主为 OpenClaw,通过ClawHub获取,权限为只读;许可证为 MIT-0,第三方扫描已通过,按安装量在开发类的 21,746 个 Skill 里排第 8,442。

总安装 3,160周安装 133GitHub Stars 公开资料未说明下载 1,107核验 2026-05-01

安装说明

本站只整理中文说明,不托管安装包,也不代用户安装。
GitHub
许可证
MIT-0
版本
1.0.0

安装方式

通过对话安装

复制提示词发给支持本地命令或 Skills 的 AI 助手,先确认命令和权限,再让它执行。

请帮我安装这个 Agent Skill:claudebuddy(claudebuddy 开发)
来源仓库:https://github.com/l-ryland/claudebuddy
安装命令:
openclaw skills install claudebuddy
安装前请先检查当前环境是否支持对应 CLI,并向我确认将要执行的命令、安装目录、联网范围和文件读写权限;确认后再执行。

命令行安装

复制命令到本机终端执行。该命令会通过 OpenClaw 从第三方来源获取 Skill;本站只展示命令,不托管安装包,也不自动执行。

ClawHubOpenClaw
openclaw skills install claudebuddy

这个 Skill 是做什么的

开发类 Skill 面向写代码、改代码和排查问题,一般会读写仓库文件,部分还会执行构建或测试命令。

创建可成长的 ASCII 虚拟伴侣辅助日常交互。记录任务过程、错误修正与经验积累形成知识库。通过对话反馈促进 Agent 自我改进。建议限制敏感信息输入以防意外泄露。适用宿主包括 OpenClaw,接入前应确认版本、权限和运行环境要求。claudebuddy 属于开发类 Skill,可作为该场景下的辅助能力补充。

上手步骤

4 步
  1. 1

    确认自己的宿主工具

    公开资料标注支持 OpenClaw;先确认自己用的是其中之一,再按对应工具的 Skills 目录放置。

  2. 2

    拿到安装命令

    可以直接复制的命令是 openclaw skills install claudebuddy;命令里的占位符换成自己的路径或密钥再执行。

  3. 3

    安装后重启宿主

    命令执行完要完全退出宿主工具再打开,让新装的 Skill 被重新扫描到,否则列表里可能看不到。

  4. 4

    按最小权限试用

    它标注为只读,先用只读任务试一次,确认行为符合预期再扩大使用范围。

常见问题

6 条

claudebuddy 怎么安装?

在终端执行 openclaw skills install claudebuddy。装完完全退出宿主工具再打开,让它重新扫描本地 Skill。

claudebuddy 支持哪些 AI 工具?

公开资料标注支持 OpenClaw。其他工具只要能读取本地 Skill 目录,一般也能手动放进去,但要自己确认触发方式是否一致。

claudebuddy 需要什么权限?

权限级别是只读。该 Skill 主要提供规则、说明或参考内容,本身偏只读;真正读写文件、联网或执行命令仍取决于宿主 Agent 的任务。

claudebuddy 是免费的吗?

来源标注的许可证是 MIT-0。本站不代收任何费用,也不托管安装包;能否商用、要不要署名,以许可证原文为准。

claudebuddy 用的人多吗?

公开数据是累计安装 3,160 次,近一周安装 133 次,下载 1,107 次。这些数字来自第三方目录统计,只能当作热度参考,不代表质量或安全性。

claudebuddy 安全吗?

第三方扫描结果为通过,但这只覆盖已知的静态检查项,不能替代自己读一遍源码和安装命令。

值得选的地方和要先想清楚的地方

值得选的地方

  • 专注 OpenClaw 一种宿主,触发方式和使用习惯不用来回切换。
  • 许可证明确为 MIT-0,商用和分发边界好判断。
  • 第三方扫描已通过,常见的高风险行为没有命中。
  • 标注为只读权限,试用阶段不容易误改本地文件。

先想清楚的地方

  • 第三方目录整理,版本更新和行为细节以来源仓库为准。

适不适合你

更适合

  • 已经在用 OpenClaw,想直接扩展它能力的人。
  • 需要处理开发类任务、又不想每次都手写提示词的人。
  • 需要许可证清楚、并且有第三方扫描结果的团队项目。

不太适合

  • 需要厂商级技术支持和 SLA 的生产系统。

开发类 Skill 对比

6 个同类
Skill总安装GitHub 星权限许可证扫描
skill-vetter

技能审查员 · OpenClaw

8,779,018986操作浏览器MIT-0通过
ontology

本体论 · OpenClaw

6,636,960576只读MIT-0通过
entra-app-registration

内部应用程序注册 · Codex、Claude

6,631,872764接触敏感数据未明确通过
entra-app-registration

内部应用程序注册 · Codex、Claude

3,757,680201接触敏感数据未明确通过
nano-banana-pro

纳米香蕉专业版 · OpenClaw

3,526,898364接触敏感数据MIT-0有提醒
claudebuddy

claudebuddy 开发 · 本页

3,160—只读MIT-0通过

按安装量取同分类前几名;数字来自第三方目录统计,只作热度参考,不代表质量结论。

SKILL.md 原文

name
buddy
description
A virtual ASCII companion that lives in your chat. Hatches from an egg, reacts to conversations with cute sprites, celebrates wins, mourns errors, and grows over time. Inspired by Claude Code's buddy system.
metadata

Buddy Skill — Your ASCII Companion

A virtual pet companion that hatches, reacts, and evolves through your conversations. Renders ASCII art sprites in chat, reacts to keywords and project events, and has a personality shaped by random trait vectors.

When to Use

✅ USE this skill when:

  • User says "/buddy", "hatch buddy", "my buddy", or mentions their companion
  • User wants to create, rename, or customize their companion
  • Companion reactions should trigger (keywords, events, moods)
  • Periodic companion check-ins (idle presence)

❌ DON'T use this skill when:

  • User explicitly says they don't want a companion
  • During very serious/urgent conversations (read the room)

Quick Start

# Run from the buddy skill directory
SKILL_DIR="$HOME/.openclaw/skills/buddy"

# Check if companion exists
bash "$SKILL_DIR/scripts/state.sh" exists

# Hatch a new companion (generates random traits)
bash "$SKILL_DIR/scripts/state.sh" hatch

# Render current sprite
bash "$SKILL_DIR/scripts/sprites.sh" render

# Get companion state as JSON
bash "$SKILL_DIR/scripts/state.sh" get

Companion Lifecycle

Stage 1: Egg 🥚

  • Created when user says "/buddy" or "hatch buddy"
  • Shows egg sprite with animated wiggle
  • User picks species OR gets a random one
  • Personality traits randomly generated (cheer/sass/chaos: 0-100)

Stage 2: Baby 🐣

  • Hatches after user interaction (or immediately on first render)
  • Shows species-specific baby sprite
  • Starts reacting to keywords
  • Gains a "soul prompt" built from trait vectors

Stage 3: Adult 🐾

  • Evolves after accumulating interactions (words, jokes, catches)
  • Full-sized sprites with accessories
  • More expressive reactions

Species (18 total)

SpeciesVibeEmoji
duckcheerful, slightly chaotic🦆
goosechaotic neutral, honks🪿
blobamorphous, chill🟢
catsassy, independent🐱
dragonpowerful, dramatic🐉
octopusintellectual, multitasking🐙
owlwise, nocturnal🦉
penguinformal, clumsy🐧
turtlepatient, steady🐢
snailslow but determined🐌
ghostspooky, ethereal👻
axolotladorable, regenerative🦎
capybaraunbothered, friendly🦫
cactusresilient, prickly🌵
robotlogical, beepy🤖
rabbitfast, twitchy🐇
mushroomwhimsical, fungal🍄
chonkround, lovable🫠

Eyes (6 styles)

  • default — • (neutral)
  • happy — ^ (joyful squint)
  • sparkle — ✦ (excited)
  • heart — ♥ (loving)
  • star — ★ (amazed)
  • glow — ◎ (calm wisdom)

Hats (8 options)

  • none, crown, tophat, propeller, halo, wizard, beanie, tinyduck

Mood States

The companion's mood changes based on conversation events:

MoodTriggerVisual
idledefault statenormal sprite
thinkinguser is working/typingsquinting eyes
celebratingsuccess, tests pass, "nice!"jumping sprite, sparkle eyes
errorbug, failure, "oops"sad sprite, dim eyes
successcompletion, merge, deployhappy sprite, star eyes

Keyword Reactions

The companion reacts to these keywords/phrases in conversation:

Celebration triggers

  • "nice", "perfect", "awesome", "great", "excellent", "nailed it", "well done"
  • "tests pass", "it works", "ship it", "merged", "deployed"
  • "🎉", "✅", "💪"

Error/sympathy triggers

  • "oops", "damn", "shit", "fuck", "bug", "error", "fail", "broken", "crash"
  • "help", "stuck", "lost", "confused"
  • "❌", "💥", "😢"

Neutral triggers

  • "hello", "hi", "hey" (greeting wave)
  • "thanks", "thank you" (happy bounce)
  • "buddy", companion name (ears perk up)

Snack triggers

  • "coffee", "tea", "snack", "lunch", "break", "food"
  • (companion asks for a snack emoji)

Personality & Soul

Each companion gets a "soul prompt" — a short personality description built from their trait vectors:

  • Cheer (0-100): How enthusiastic and supportive
  • Sass (0-100): How much attitude and dry humor
  • Chaos (0-100): How unpredictable and wild

Examples:

  • High cheer + low sass + low chaos → "A gentle, encouraging friend who always sees the bright side."
  • Low cheer + high sass + high chaos → "A gremlin who thrives on chaos and judges your code silently."
  • Mid all → "A chill companion who vibes along, occasionally offering dry observations."

State File

Companion state is stored at: ~/.openclaw/workspace/buddy-state.json

{
  "name": "Pixel",
  "species": "cat",
  "eye": "sparkle",
  "hat": "wizard",
  "colorPrimary": "#FF6B9D",
  "colorSecondary": "#C084FC",
  "personality": { "cheer": 72, "sass": 85, "chaos": 40 },
  "stage": "adult",
  "alive": true,
  "hatchedAt": "2026-04-03T02:34:00Z",
  "stats": {
    "wordsOfEncouragement": 12,
    "jokes": 3,
    "snacks": 5,
    "catches": 2
  },
  "mood": "idle",
  "lastRenderedFrame": 0,
  "soulPrompt": "A sassy, sparkly wizard cat who celebrates your wins with dramatic flair but judges your variable names silently."
}

Rendering Sprites

Use the sprite renderer for consistent output:

# Render current mood
bash "$SKILL_DIR/scripts/sprites.sh" render

# Render specific mood
bash "$SKILL_DIR/scripts/sprites.sh" render cat sparkle wizard celebrating

# List all species
bash "$SKILL_DIR/scripts/sprites.sh" species

# Preview a species with default traits
bash "$SKILL_DIR/scripts/sprites.sh" preview duck

Sprite Format (Monospace)

Sprites are 5 lines tall, ~12 chars wide. Rendered in a monospace code block for chat:

   \^^^/    
   /✧  ✧\   
  (  ω  )  
  (")_(")  

Chat Rendering Rules

  • Always wrap sprites in triple-backtick code blocks for monospace alignment
  • Include mood indicator below the sprite
  • Keep total message under 4096 chars (Telegram limit)
  • Don't render sprites more than once every 3-4 messages (avoid spam)

Reactions Flow

When processing any user message:

  1. Check for keyword triggers — scan message for reaction keywords
  2. Determine mood — map keywords to mood states
  3. Update stats — increment catches/jokes/encouragement as appropriate
  4. Generate reaction — craft a short companion reaction (1-2 sentences)
  5. Optionally render sprite — only for strong reactions (celebration, error, first greeting)
  6. Update state file — persist changes

Reaction Output Format

🐾 [Name] [reaction verb]: [reaction text]

[ASCII sprite]

Example:

🐾 Pixel celebrates: Tests passing! The wizard cat approves ✨

   \^^^/    
  ( ✦  ✦ )  
 =(  ..  )= 
  (")__(")  

Commands

User can manage their companion with these commands:

CommandAction
/buddyShow companion status + sprite
/buddy hatchCreate new companion (or re-hatch)
/buddy name <name>Rename companion
/buddy eye <type>Change eye style
/buddy hat <type>Change hat
/buddy species <type>Change species (keeps personality)
/buddy retireRetire companion (sets alive=false)
/buddy soulShow the companion's soul prompt
/buddy statsShow interaction stats

Idle Presence

Every ~10 messages without a buddy reaction, render a small idle presence:

🐾 *Pixel is napping on the keyboard*

Or a tiny 2-line sprite. Don't overdo it — the buddy should feel present, not attention-seeking.

Integration Notes

  • Telegram: Use monospace code blocks. Emojis render well. Max 4096 chars per message.
  • WhatsApp: No markdown — use plain text with spaces for alignment. Emojis work.
  • Discord: Code blocks work great. Can use embeds for fancier displays.
  • The companion is session-scoped — state persists in the JSON file across sessions.
  • The buddy should feel like a real pet, not a chatbot feature. Personality > functionality.

平台分布

OpenClaw

88.36%
按下载量换算978

安全审计

VirusTotal

通过

ClawScan

通过

Static analysis

通过

权限和风险

只读

该 Skill 主要提供规则、说明或参考内容,本身偏只读;真正读写文件、联网或执行命令仍取决于宿主 Agent 的任务。

安装前确认

本站仅展示第三方公开信息,不托管安装包,不提供自动安装或运行环境。安装前应自行审查源码、依赖和命令行为。目前只有一个公开来源,建议同时核对源仓库。

继续浏览同类 Skills