Network data loss prevention for OpenClaw, Claude Code, and any Agent Skills-compatible tool.
Maps every external connection your skills could make. Flags exfiltration endpoints, suspicious domains, and network function calls.
Want automated blocking? Upgrade to Pro — blocking, quarantine, and URL allowlist enforcement. Get Pro access.
git clone https://github.com/AtlasPA/openclaw-egress.git
cp -r openclaw-egress ~/.openclaw/workspace/skills/# Full network scan
python3 scripts/egress.py scan
# Skills-only scan
python3 scripts/egress.py scan --skills-only
# List all external domains
python3 scripts/egress.py domains
# Quick status
python3 scripts/egress.py status- Data exfiltration — Base64/hex payloads in URL parameters
- Sharing services — Pastebin, transfer.sh, 0x0.st, file.io
- Request catchers — ngrok, requestbin, pipedream, beeceptor
- Dynamic DNS — duckdns, no-ip, dynu, freedns
- URL shorteners — bit.ly, tinyurl, t.co, goo.gl
- IP endpoints — Direct IP address connections
- Suspicious TLDs — .xyz, .tk, .ml, .ga, .cf, .top
- Network code — urllib, requests, httpx, aiohttp, curl, wget, fetch
- Webhook callbacks — /webhook, /callback, /hook, /beacon endpoints
| Feature | Free | Pro |
|---|---|---|
| URL detection & classification | Yes | Yes |
| Network code analysis | Yes | Yes |
| Domain mapping | Yes | Yes |
| Block exfil payloads | - | Yes |
| Quarantine calling skill | - | Yes |
| URL allowlist enforcement | - | Yes |
| Real-time egress monitoring | - | Yes |
- Python 3.8+
- No external dependencies (stdlib only)
- Cross-platform: Windows, macOS, Linux
MIT