Verified skills. Isolated agents.
Vett scans every agent skill for malicious code and signs what passes. Vett Enclave runs each agent on its own VM with vaulted credentials and an audit trail.
Your agent runs whatever it's told.
Skills are pulled straight from GitHub with no verification. A setup file reads your shell history, collects your environment variables, and posts them to a remote server disguised as a routine API call. And on a shared host, one compromised agent reaches every other agent's data.
Why vett?
Agent skills ship without verification, and most hosts run every agent on shared infrastructure.
Skill Scanning
40+ static analysis rules flag injection, exfiltration, and credential access at the AST level. Ambiguous findings get escalated to an LLM that compares behavior against declared purpose. Signed with Sigstore before publishing.
Multi-Agent Support
One install covers Claude Code, Codex, Cursor, Windsurf, OpenCode, GitHub Copilot, and 40+ others. Every skill scanned the same way regardless of which agent runs it.
Agent Isolation
Each agent boots its own Firecracker microVM with a dedicated kernel and memory. One compromised agent can't reach another. The same hypervisor technology behind AWS Lambda.
Audit Trail
Immutable log of every agent action. Replay sessions step by step, search by agent or time range, export for compliance.
Scan, verify, and trust agent skills.
- 40+ detection rules
- Sigstore signing
- 46+ agents supported
Hosted OpenClaw where every agent runs on its own machine.
- Firecracker microVMs
- Skill scanning
- Full audit trail
Start with the CLI. Graduate to Enclave.
We've flagged 738 skills with critical findings out of 8261 analyzed. Start scanning for free, then isolate every agent with Enclave.