openclaw-lead-search

Self-configuring autonomous lead search agent for OpenClaw.

Describe what you want to find in plain text. The agent generates all configs, searches the web, validates and deduplicates results, scores leads by priority, and improves its own strategies over time.

Quick Start

git clone https://github.com/user/openclaw-lead-search.git
cd openclaw-lead-search
./setup.sh

The setup wizard asks for:

1. A skill name (e.g., my-leads)
2. A text description of what you're looking for

That's it. On the first run, the agent reads your description and auto-generates all config files (schema, scoring rules, search strategies, validators, etc.).

How It Works

Bootstrap (first run): Agent reads your brief + config format reference, generates 9 config files, validates them, and starts searching.

Operational (every run): Standard pipeline: load memory → choose sources → search → validate → deduplicate → merge → score → export → update memory

Self-improve (every ~10 runs): Agent analyzes its own performance data and updates configs — removes dead sources, adds new ones, tunes scoring, expands blacklist.

Structure

scripts/    12 generic pipeline scripts (same for any domain)
config/     Generated configs + reference docs
output/     Runtime data (DB, memory, CSV export, backups)

Requirements

• macOS with OpenClaw installed
• Python 3.9+
• crawl4ai skill (for web scraping)

License

MIT