Pattern Detector

Vetted and production-tested agent skill.

Pattern Detector

**Category:** Intelligence

**Badge:** Vetted

**Author:** AgentVet

**Version:** 1.0

---

Overview

Pattern Detector scans your Zep knowledge graph for recurring entity clusters and converging patterns. It finds when 3+ sources confirm the same signal, detects emerging themes before they become obvious, and surfaces cross-domain connections.

How It Works

1. Queries Zep graph for entity clusters and connection density

2. Analyzes week-over-week changes in graph structure

3. Detects when topics accumulate enough data points to confirm a pattern (3+ sources = candidate, 10+ = confirmed)

4. Surfaces converging themes across different conversation threads

5. Saves results to `compound-outputs/patterns/` for compounding intelligence

Installation

# Requires Zep Cloud (Flex plan) and OpenClaw
pip install zep-cloud

# Copy the pattern detector script
cp scripts/zep_pattern_detector.py /path/to/workspace/scripts/

# Run manually
python3 scripts/zep_pattern_detector.py

Cron Setup (OpenClaw)

Add to your HEARTBEAT.md or OpenClaw cron:

{
  "name": "pattern-detector",
  "schedule": { "kind": "cron", "expr": "0 16 * * 0", "tz": "UTC" },
  "payload": {
    "kind": "agentTurn",
    "message": "Run the pattern detector: python3 scripts/zep_pattern_detector.py. Report patterns found."
  }
}

What You Get

• Weekly pattern detection across research and conversations
• Finding converging signals from multiple sources
• Cross-domain connection discovery
• Emerging trend identification before they hit mainstream
• Graph health metrics (connection density, orphan entities, growth rate)

Configuration

Requirements

• Python 3.10+
• zep-cloud SDK
• OpenClaw (for cron scheduling)
• Zep Cloud Flex plan

License

MIT — Free to use, modify, and distribute.

Get the full skills pack

Enter your email to download all 15 AgentVet skills as a pack. No spam, just occasional updates on new skills.