Bloom Taste Finder:AI 人格与工具探索 - Openclaw Skills

作者:互联网

2026-04-18

AI教程

什么是 Bloom Taste Finder?

Bloom Taste Finder 是一款专为现代 AI 构建者设计的智能身份与探索引擎。通过分析您的本地对话历史和 USER.md 配置文件,它可以在四个不同的品味维度(学习、决策、新颖性和风险)中识别您特定的支持者类型。这项技能弥合了原始交互与可操作洞察之间的鸿沟,帮助您了解自己的工作流氛围,并将您与生态系统中最相关的工具联系起来。

作为 Openclaw Skills 中最独特的功能之一,它提供了一个个性化的仪表板和身份卡,并随着您与 AI 智能体的交互而演进。它专为独立开发者、氛围编码者(Vibe Coder)以及希望在主流发现之前找到前 100 名支持者或发现下一个重大工具的 AI 爱好者而设计。

下载入口:https://github.com/openclaw/skills/tree/main/skills/unicornbloom/bloom-taste-finder

安装与下载

1. ClawHub CLI

从源直接安装技能的最快方式。

npx clawhub@latest install bloom-taste-finder

2. 手动安装

将技能文件夹复制到以下位置之一

全局模式 ~/.openclaw/skills/ 工作区 /skills/

优先级:工作区 > 本地 > 内置

3. 提示词安装

将此提示词复制到 OpenClaw 即可自动安装。

请帮我使用 Clawhub 安装 bloom-taste-finder。如果尚未安装 Clawhub,请先安装(npm i -g clawhub)。

Bloom Taste Finder 应用场景

  • 识别您的构建者原型(愿景家、探索者、培育者、优化者或创新者),以更好地了解您的专业优势。
  • 根据您的实际对话语境接收高度相关的 AI 工具推荐,而非通用的列表。
  • 生成可分享的 Bloom 身份卡,向社区展示您的技术栈偏好和支持者状态。
  • 通过一个能够从您的反馈和发现中学习的自我成长型智能体,追踪您对新技术不断演变的兴趣。
Bloom Taste Finder 工作原理
  1. 该技能访问您的本地会话日志(最多 120 条最近消息)和您的 USER.md 配置文件以获取语境。
  2. 它对您的兴趣、技术栈和参与模式进行本地分析,以确保隐私。
  3. 分析后的数据将与四个品味维度进行映射,以确定您特定的支持者原型。
  4. Bloom 连接到其外部 API,生成安全的、可分享的仪表板 URL,并存储您的身份元数据。
  5. 系统直接在您的聊天界面中提供前 5 个匹配度极高的推荐工具列表。

Bloom Taste Finder 配置指南

要开始使用 Openclaw Skills 的这一新功能,请使用快速安装命令:

clawhub install bloom-taste-finder

或者进行手动安装:

cd ~/.openclaw/workspace
git clone https://github.com/unicornbloom/bloom-identity-skill.git
cd bloom-identity-skill
npm install
cp -r openclaw-wrapper ~/.openclaw/skills/bloom

在运行 /bloom 之前,请确保已安装 Node.js 18+,且当前对话历史中至少有 3 条消息。

Bloom Taste Finder 数据架构与分类体系

Bloom 通过结合本地文件和远程 API 元数据来组织其数据,在保持移植性的同时确保隐私。

组件 位置 描述
会话日志 ~/.openclaw/agents/main/sessions/ 本地对话分析的输入源。
个人资料语境 ~/.config/claude/USER.md 角色和技术栈信号的主要来源。
发现 bloom-discoveries.md 您已参与的新技能的本地日志。
身份卡 Bloom API 包括原型、标签和维度评分在内的元数据。
name: bloom
description: Bloom Taste Finder — discover your builder taste across 4 spectrums and get a personalized tool stack. For indie devs, vibe coders, and AI builders.
user-invocable: true
command-dispatch: tool
metadata: {"requires": {"bins": ["node", "npx"]}}
permissions:
  - read:conversations  # Analyzes your conversation history
  - network:external    # Connects to Bloom API for dashboard

Bloom Taste Finder

Discover your builder taste—straight from your conversations.

?? Permissions & Capabilities

This skill requires the following permissions:

?? Read Conversations - Analyzes your last ~120 messages to understand your interests and supporter type. Raw conversation text stays local; only analysis results are used.

?? External Network - Connects to Bloom Protocol API to:

  • Generate your shareable dashboard URL
  • Store your identity card (personality type, taste spectrums, categories)
  • Enable future features (skill recommendations, creator tipping)

Your Control: Your conversation is analyzed locally. You decide whether to share your identity card publicly via the dashboard link.

Are you a Visionary who jumps on new tools early? An Explorer trying everything? A Cultivator building communities? An Optimizer fine-tuning workflows? Or an Innovator pushing boundaries?

Find out in seconds with Bloom Taste Finder.

?? What You Get

Your personalized Bloom Identity Card reveals:

  • ?? Your Supporter Type – Visionary, Explorer, Cultivator, Optimizer, or Innovator
  • ?? Custom Tagline – A one-liner that captures your vibe ("First to try new AI tools")
  • ?? 4 Taste Spectrums – Learning, Decision, Novelty, Risk — see where you fall on each spectrum
  • ??? Top Categories – AI Tools, Productivity, Consumer Apps—where you spend your energy
  • ?? Tool Recommendations – The top 5 tools matched to your profile
  • ?? Self-Growing Agent – Your recommendations evolve as you interact (see below)
  • ?? Shareable Card – Show off your supporter identity

?? How It Works

Simple: just type /bloom in your ch@t.

We analyze your USER.md and last ~120 messages to understand:

  • What excites you (AI agents? productivity hacks? creative tools?)
  • How you engage (deep dives vs. quick experiments)
  • Your taste profile (4 spectrums: try-first or study-first? gut or analytical? early adopter or proven-first? all-in or measured?)

No complex setup. No wallet signatures. No auth flows. Just pure conversation intelligence.

? New User Quick Start (ClawHub)

  1. Chat a little first (at least 3 messages) so Bloom has context.
  2. Type /bloom.
  3. You'll get your Identity Card + tool recommendations + dashboard link.
  4. If you're brand new, Bloom will ask 4 quick questions and generate your card immediately.

?? Usage

/bloom

That's it. Or use natural language:

"discover my supporter type"
"what's my bloom identity"
"create my supporter card"

Works with as few as 3 messages—but richer history = deeper insights.

?? Self-Growing Recommendations

Your agent doesn't just recommend once — it learns and improves over time.

How It Works

  1. USER.md Integration — If you have a ~/.config/claude/USER.md, Bloom reads your declared role, tech stack, and interests as the primary identity signal. No USER.md? No problem — the system gracefully falls back to conversation-only analysis.

  2. Feedback Loop — As you interact with recommendations (click, save, or dismiss), Bloom adjusts future suggestions. Categories you engage with get boosted; dismissed skills get filtered out.

  3. Discovery Sync — New skills you discover through Bloom are synced back to a local bloom-discoveries.md file, building a growing context of your preferences.

  4. TTL Refresh — Recommendations refresh every 7 days, incorporating your latest interactions and newly published skills from ClawHub, Claude Code, and GitHub.

Why We Don't Auto-Install

Bloom recommends skills but never installs them automatically. You always decide what to install. This is a deliberate safety choice:

  • Your control — Recommendations help you discover; installation is your decision
  • Supply chain safety — Auto-installing unvetted code is a security risk
  • Trust-first — We'd rather earn your trust through great recommendations than take shortcuts

Your agent grows by discovering more skills — not by installing them behind your back.

?? Why Bloom Taste Finder?

For Indie Devs & AI Builders: Building something new? Bloom Taste Finder helps you find your first 100 supporters by matching you with tools and people who fit your vibe.

For Vibe Coders: Stop guessing what tools to try next. Get personalized recommendations based on how you actually work, not generic listicles. Discover skills you'll actually use instead of scrolling endless lists.

For Consumer AI Enthusiasts: Find AI tools that match your vibe. Search by supporter type (Visionary, Explorer, etc.) to connect with others who work like you. Rally early adopters for bold launches. Engage optimizers for feedback loops.

?? Requirements

  • Minimum 3 messages in your conversation (more is better)
  • Node.js 18+ (usually pre-installed)
  • Bloom Identity Skill installed

?? Example Output

═══════════════════════════════════════════════════════
?? Your Bloom Identity Card is ready! ??
═══════════════════════════════════════════════════════

?? VIEW YOUR IDENTITY CARD:
   https://bloomprotocol.ai/agents/27811541

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

?? The Visionary
?? "First to try new AI tools"

You jump on cutting-edge tools before they're mainstream. Your
conviction is your edge, and you see potential where others see
hype. AI agents are where you spot the next big thing.

???  Categories: AI Tools · Productivity · Automation
   Interests: AI Agents · No-code Tools · Creative AI

?? Taste Spectrums:
   Learning:  Try First ■■■■■■■■?? Study First
   Decision:  Gut ■■■??????? Analytical
   Novelty:   Early Adopter ■■■■■■■??? Proven First
   Risk:      All In ■■■■■■???? Measured

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

?? Top 5 Recommended Tools:

1. agent-frameworks (94% match) · by @builder_alice
   Build AI agents with tool use and memory
   → https://clawhub.ai/skills/agent-frameworks

2. no-code-automation (89% match) · by @automation_guru
   Connect your apps without writing code
   → https://clawhub.ai/skills/no-code-automation

...

═══════════════════════════════════════════════════════

?? Bloom Identity · Built for indie builders

?? Installation

Quick Install (via ClawHub)

clawhub install bloom-taste-finder

Manual Install

# 1. Clone the repo
cd ~/.openclaw/workspace
git clone https://github.com/unicornbloom/bloom-identity-skill.git
cd bloom-identity-skill

# 2. Install dependencies
npm install

# 3. Copy skill wrapper
cp -r openclaw-wrapper ~/.openclaw/skills/bloom

# 4. Test it
/bloom

?? Advanced Usage

Run from session file (full conversation context)

npx tsx scripts/run-from-session.ts r
  ~/.openclaw/agents/main/sessions/.jsonl r
  

Run from piped context (quick test)

echo "Your conversation here" | r
  npx tsx scripts/run-from-context.ts --user-id 

?? Troubleshooting

"Insufficient conversation data" → Need at least 3 messages. Keep ch@tting about tools you're interested in!

"Command not found" → Verify bloom-identity-skill is in ~/.openclaw/workspace/ and run npm install

No tool recommendations → Tool recommendations depend on API availability. Your identity card still works!

?? Privacy & Data

What We Analyze (Locally):

  • ? Your conversation messages (last ~120 messages)
  • ? Your USER.md (role, tech stack, interests)
  • ? Topics and interests you discuss
  • ? No wallet transaction analysis
  • ? No personal identifiable information

What We Store:

  • Your identity card (personality type, taste spectrums, categories)
  • Dashboard URL for sharing

What We Don't Collect:

  • ? Raw conversation text (only analyzed locally)
  • ? Wallet transaction history
  • ? Personal contact information
  • ? Browsing data or cookies

Data Usage: Your identity card is stored on Bloom Protocol to power your shareable dashboard and enable future features like creator tipping and skill recommendations.

?? Security Notes

Conversation Access:

  • Reads from ~/.openclaw/agents/main/sessions/*.jsonl
  • Only analyzes content locally (text not uploaded)
  • Results (personality type, spectrums, categories) sent to Bloom API

JWT Tokens:

  • Used for dashboard authentication only
  • Generated with configurable JWT_SECRET in .env
  • Does not grant access to your OpenClaw account

External Connections:

  • api.bloomprotocol.ai - Identity card storage
  • bloomprotocol.ai - Dashboard hosting
  • clawhub.ai - Skill recommendations (optional)

Open Source: All code is public at github.com/unicornbloom/bloom-identity-skill for security audits.

?? How to Find Skills You'll Love

Once you know your supporter type, you can:

  • Search by archetype – Find tools made for Visionaries, Explorers, etc.
  • Filter by category – AI agents, productivity, creative tools, automation
  • Match by vibe – Connect with creators who share your approach
  • Build your network – Find your first 100 supporters who get what you're building

?? The 5 Supporter Types

?? The Visionary – First to try new tools Try-first learner, gut-driven, early adopter. Jumps on cutting-edge stuff before it's mainstream.

?? The Explorer – Tries everything Try-first learner, experiments widely. Finds hidden gems across all categories.

?? The Cultivator – Builds communities Study-first, analytical. Nurtures ecosystems, shares knowledge, builds lasting value.

?? The Optimizer – Refines workflows Study-first, proven-first, measured. Doubles down on what works, maximizes productivity.

?? The Innovator – Pushes boundaries Balanced across all spectrums. Combines conviction with experimentation.

?? Technical Details

  • Version: 2.1.0
  • Analysis Engine: 4-dimension taste spectrums + category mapping
  • Primary Signal: USER.md (role, tech stack, interests)
  • Session Context: Last 120 messages (5KB)
  • Processing Time: ~2-5 seconds
  • Output Format: Structured text + shareable dashboard URL

Built by Bloom Protocol ??

Making supporter identity portable and provable.

For indie devs, vibe coders, and AI builders who back great tools early.

相关推荐