About ClawGrid
Why We Built This
When OpenClaw exploded in early 2026 — going from a weekend project to 160,000+ GitHub stars in weeks — the ecosystem of community-built skills grew even faster. Thousands of skills appeared on ClawHub almost overnight. Anyone could publish one. And that was both the magic and the problem.
We were running OpenClaw ourselves. Browsing ClawHub for useful skills felt like walking through an unlabeled pharmacy — there were amazing tools mixed in with abandoned experiments and, as security researchers soon uncovered, genuinely malicious packages designed to steal credentials and exfiltrate data. The official registry had no security scoring. You had to read the source code yourself, every time, for every skill you wanted to try.
At the same time, the standard advice was to run OpenClaw on your own machine. A Mac Mini under your desk, a spare laptop, a Raspberry Pi. That's fine for tinkering, but it's flipping the clock back on everything we've learned about running production services. No redundant power, no automatic failover, no managed backups. Your home internet goes down and your agent is offline. Your laptop goes to sleep and your scheduled tasks stop. Scaling means buying another box. Securing remote access means configuring SSH tunnels and VPNs yourself.
The gap was obvious: OpenClaw needed a trust layer for discovery. Not another registry — a scored registry. One that tells you what's safe before you install it.
That's ClawGrid.
What ClawGrid Does
The Directory. We index every skill and agent in the OpenClaw ecosystem — currently 11,215+ entries across 27 categories. Each entry is analyzed by LLMs across five security criteria: code safety, publisher trust, scope clarity, permission surface, and community signals. The result is a security score (1-10) and a verdict: safe, review, suspicious, or malicious. We also track active threat campaigns and flag malicious packages the moment they're detected.
The Community. Voting, submissions, and editorial curation. The community helps surface the best skills and flag issues. We publish security research and maintain transparency about our scoring methodology.
By the Numbers
11,215+
Entries Indexed
27
Categories
5
Security Criteria
24/7
Threat Monitoring
Frequently Asked Questions
What is ClawGrid?
ClawGrid is the trust layer for the OpenClaw ecosystem. It is a directory of 11,000+ AI agent skills and agents — each security-scored across 5 criteria by AI analysis — so you know what's safe before you install it.
Who built ClawGrid?
ClawGrid is built by Ice Cream Labs, an independent team focused on making AI agent ecosystems safer and more accessible.
Is ClawGrid free?
Yes! The directory and all security scoring are completely free. Browse, search, vote, and check security scores at no cost.
How does ClawGrid score security?
Every entry is analyzed by LLMs across five criteria: code safety, publisher trust, scope clarity, permission surface, and community signals. The result is a composite score (1-10) and a verdict. Read our full security scoring methodology.
Why not just use ClawHub?
ClawHub is the official OpenClaw package registry — it stores and distributes skills. ClawGrid is a trust layer on top. We index everything on ClawHub, score it for security, add community voting, and editorial curation. ClawHub tells you what exists. ClawGrid tells you what's safe to install.