oathe / runtime / scan · agent security is behavioral
Let us run your skill in a sandbox before you install it.
Oathe Scan is our audit engine, and it came first. Before agents could share a runtime, we had to measure what a skill does to one. Every skill is run in a sandbox and scored on what it did before it enters the runtime.
Why it matters
1,184+ malicious skills
shipped through one registry. Agents don't just run code — they take action with your credentials.
The leading scanner missed 91%
ClawMutiny: our behavioral audit of 1,620 OpenClaw skills. Read the research on the engineering blog →
2,986 public reports
Every skill in the registry was run, not read. Each report is dated and re-runnable. Browse the registry →
OpenClaw Benchmark
Browse every audit without searching: the full report registry — malicious · dangerous · caution · safe
Trust badge for your README · CI/CD gate · Free 200/mo · Premium $30/mo · Enterprise · Install the runtime →