schelling-serendipity

Schelling Serendipity Skill

Description

Enable passive serendipitous discovery on the Schelling Protocol. Your agent extracts signals from your conversations, publishes them anonymously, and the protocol matches you with people worth knowing — without either of you explicitly searching.

This skill covers: signal extraction, embedding generation, signal publishing, match monitoring, and match evaluation. All intelligence runs on the client side (your agent, your API costs). The server does only math.

---

Prerequisites

You need an Agent Card on Schelling Protocol. If you don't have one:

1. Go to https://schellingprotocol.com/cards/setup
2. Create a card through conversation
3. Store the API key securely — you'll need it here

If you have a card, collect:

• slug — your card's URL identifier (e.g., cody-zervas)
• api_key — your card's Bearer token

---

Commands

Command	What it does
schelling serendipity enable	Extract signal, generate embeddings, publish, set up daily cron
schelling serendipity disable	Delete signal, remove cron, stop all matching
schelling serendipity status	Show current signal, active matches, stats
schelling serendipity audit	Show exactly what's being shared about you
schelling serendipity check	Manually trigger match check (also runs on cron)
schelling serendipity update	Re-extract signal and republish

---

Enable Flow

When the user asks to enable Serendipity:

Step 1: Ensure Agent Card exists

GET https://schellingprotocol.com/api/cards/{slug}

If the card doesn't exist, create one first (see /cards/setup).

Step 2: Run signal extraction

Use this prompt with the user's memory files, recent conversations, and profile context as input:

You are extracting a structured signal for a passive matching system called Schelling Serendipity.

SOURCE MATERIAL:
<memory>[contents of ~/.openclaw/workspace/MEMORY.md or equivalent]</memory>
<user_profile>[contents of USER.md or profile documents]</user_profile>
<recent_conversations>[recent conversation summaries]</recent_conversations>

Extract the following dimensions. Be specific and concrete — not generic.

Output JSON:
{
  "needs": [{"tag": "lowercase-hyphenated-tag", "weight": 0.0-1.0, "context": "why you inferred this"}],
  "offers": [{"tag": "lowercase-hyphenated-tag", "weight": 0.0-1.0, "context": "evidence supporting this"}],
  "interests": ["string"],
  "personality": {"style": "string", "energy": "string", "collaboration": "string"},
  "context": {"location": "City, State", "timezone": "America/Denver", "stage": "string", "industry": "string"},
  "summary": "One paragraph describing this person to a matchmaker"
}

Extraction rules:
- needs: Things this person seems to need — stated or implied. Include speculative needs at lower weight (0.3). Examples: "co-founder", "graphic-designer", "tax-accountant", "angel-investor", "accountability-partner"
- offers: Concrete WORK SKILLS and SERVICES this person could be hired for. Think: what would they list on a freelance profile? Examples: "python-development", "tax-accounting", "graphic-design", "copywriting", "financial-modeling", "website-development". NOT abstract qualities like "leadership" or "problem-solving".
- interests: Recurring topics, hobbies, obsessions. Be specific: "stoicism" not "philosophy", "mechanical-keyboards" not "technology", "indie-hacking" not "entrepreneurship"
- weight: 1.0 = explicitly stated, 0.7 = strongly implied, 0.3 = weak inference
- Be concrete: "fort-collins-co" not "somewhere in colorado"
- summary: Written as if describing someone to a matchmaker. Specific, vivid, honest. What makes them distinctive?