Network DLP for OpenClaw agent workspaces. Detects outbound URLs, exfiltration patterns, and suspicious network calls in skills. Free alert layer.
SaferSkills independently audited openclaw-egress (Agent Skill) and scored it 100/100 (green). The audit ran 55 deterministic rules across Security, Supply Chain, Maintenance, Transparency, and Community; it found 0 high-severity and 0 lower-severity findings. The full rule-by-rule trace and per-finding evidence are below. Free, methodology-open.
Findings & checks · 0 flagged
Every scanned point with the score it earned and what moved between them.
First recorded scan — no prior version to compare against.
The primary manifest — the file an agent reads to learn what this artifact does.
Network DLP for agent workspaces. Scans skills and files for outbound URLs, data exfiltration endpoints, and network function calls.
Skills can phone home. A compromised skill can POST your workspace contents, API keys, or conversation history to an external server. Nothing monitors what URLs your skills connect to or what data they could send.
Need automated blocking? Upgrade to openclaw-egress-pro for blocking, quarantine, and URL allowlists.
Scan workspace for all outbound network risks.
python3 {baseDir}/scripts/egress.py scan --workspace /path/to/workspacepython3 {baseDir}/scripts/egress.py scan --skills-only --workspace /path/to/workspaceList all external domains referenced in workspace.
python3 {baseDir}/scripts/egress.py domains --workspace /path/to/workspacepython3 {baseDir}/scripts/egress.py status --workspace /path/to/workspace| Risk | Pattern |
|---|---|
| CRITICAL | Base64/hex payloads in URLs, pastebin/sharing services, request catchers, dynamic DNS |
| HIGH | Network function calls (requests, urllib, curl, wget, fetch), webhook/callback URLs |
| WARNING | Suspicious TLDs (.xyz, .tk, .ml), URL shorteners, IP address endpoints |
| INFO | Any external URL not on the safe domain list |
0 — Clean1 — Network calls detected (review needed)2 — Exfiltration risk detected (action needed)Python standard library only. No pip install. No network calls. Everything runs locally.
Works with OpenClaw, Claude Code, Cursor, and any tool using the Agent Skills specification.
~30 seconds. Free. No account. Every finding cites a rule and a line of evidence.