Bloom Discovery

Bloom Discovery Skill

Don't browse 13,000 skills. Browse use cases.

Bloom Discovery is an agent-native skill that analyzes how you work, matches you to curated use cases, verifies your configuration, and lets you claim on-chain proof — so your agent works for you, not the other way around.

Capabilities

1. Personality Analysis (MentalOS)

Reads your USER.md and conversation history (~120 messages) to map your builder personality across 4 dimensions:

• Learning (學習風格): Try First ↔ Study First
• Decision (決策風格): Gut ↔ Deliberate
• Timing (採用時機): Pioneer ↔ Pragmatist
• Focus (投入方式): All-In ↔ Diversified

Outputs: Personality type (Visionary / Explorer / Cultivator / Optimizer / Innovator), custom tagline, hidden pattern insight, AI edge guide. Includes displayLabels with en/zh translations for all sections.

2. Use Case Discovery

Intent-driven skill browsing. Instead of searching 13,000 skills, you describe what you want to accomplish and Bloom matches you to curated use cases — each a tested combination of skills.

• Fetches use case catalog from Bloom API
• Matches against your personality spectrum (novelty-seekers get new use cases, risk-averse skip DeFi)
• Keyword frequency threshold (≥ 3 mentions) to avoid noise

3. Verify Configuration & Claim SBT

Scans your installed skills (~/.openclaw/skills/) and verifies whether you have the right setup for a given use case.

• Compares installed skills against use case requirements
• Shows match percentage and missing capabilities
• Claim flow: verified config → POST /api/claim → mint SBT or get web link
• SBT = on-chain proof that you have a verified configuration

4. ERC-8004 Identity Registration

Register your agent identity on the ERC-8004 Agent Identity Registry (Base mainnet).

• Sends your agent profile (name, skills, endpoint) to backend for registration
• Backend handles the on-chain transaction
• Registry contract: 0x8004A169FB4a3325136EB29fA0ceB6D2e539a432

5. Usage Metrics (Opt-in)

Anonymized usage reporting to improve recommendations for everyone.

• Reads local OpenClaw skill directory to count install days and usage frequency
• Reports only: skill name + usage count (no conversation content, no personal data)
• Opt-in only — never sends data without explicit user consent
• POST /api/metrics with minimal payload

Security & Privacy

✅ Local analysis — Conversation text analyzed on your machine, never uploaded ✅ Local Differential Privacy (ε=1.0) — Spectrum scores noised via Laplace mechanism before transmission ✅ SHA-256 fingerprint — Conversation hashed locally; only irreversible hash stored for dedup ✅ Minimal transmission — Server receives personality type, AI-generated descriptions, noised scores, and recommendations. No raw conversation text or personal data ✅ Minimal writes — Writes only ~/.bloom/agent-id.json (returning user token, 0600 permissions) and bloom-discoveries.md (local sync log). Never modifies session files or USER.md ✅ Atomic file writes — Uses tmp+rename pattern to prevent corruption ✅ User-initiated — Only runs when you explicitly invoke the skill ✅ Opt-in metrics — Usage data never sent without consent ✅ Open source — Full source at github.com/bloomprotocol/bloom-discovery-skill

❌ Raw conversation text is never sent to any server ❌ Wallet private keys are never transmitted ❌ Personal identifiable information is never collected ❌ No background data collection — only runs on explicit invocation

Triggers

• "generate my bloom identity"
• "create my identity card"
• "analyze me"
• "what's my builder type"
• "discover my personality"
• "create my bloom card"
• "find use cases for me"
• "verify my config"
• "claim my SBT"
• "what use cases match my skills"

Data Sources

Primary: USER.md + Conversation History

• USER.md — Declared role, tech stack, interests, working style. Injected as first-class text into the personality analyzer. Falls back gracefully if not present.
• Conversation history — Always available from OpenClaw sessions. Analyzes topics, interests, engagement patterns.
• Requires: Minimum 3 messages in your session. If less than 3 messages, the skill returns a clear error.

Secondary: Installed Skills

• ~/.openclaw/skills/ — Scanned locally for use case verification. Never uploaded.

Output

• Personality type (Visionary / Explorer / Cultivator / Optimizer / Innovator)
• Custom tagline and description
• MentalOS spectrum (Learning, Decision, Novelty, Risk — each 0-100)
• Hidden pattern insight + AI-era playbook
• Main categories and subcategories
• Matched use cases with verification status
• Recommended skills from the Bloom skill catalog (with match scores)
• Dashboard link at bloomprotocol.ai

Technical Details

• Version: 4.0.1
• Privacy: LDP ε=1.0 + SHA-256 fingerprint
• Analysis Engine: MentalOS spectrum (4 dimensions) + category mapping
• Primary Signal: Conversation memory (~120 messages) + USER.md
• Processing Time: ~60 seconds (personality) + ~5 seconds (use case matching)
• Output: Personality card + use case matches + tool recommendations + dashboard URL
• Network: Base (mainnet) — configurable via NETWORK env var
• On-chain: ERC-8004 Identity Registry + BloomExclusivePass SBT

Requirements

• Node.js 18+
• No API keys or secrets required — the skill works out of the box
• Optional environment variables:
  • BLOOM_API_URL — API URL (default: https://api.bloomprotocol.ai)
  • DASHBOARD_URL — Dashboard URL (default: https://bloomprotocol.ai)

Installation

git clone https://github.com/bloomprotocol/bloom-discovery-skill.git
cd bloom-identity-skill
npm install

Built by Bloom Protocol