How Crawdad addresses known threats to AI agents.
| Threat | Status | Details |
|---|---|---|
| CVE-2026-25253 (OpenClaw RCE) | Mitigated | Skill attestation + inbound firewall |
| LiteLLM supply chain (March 2026) | Partially | Skill attestation + outbound scanning |
| MCP prompt injection | Mitigated | Semantic firewall + policy engine |
| RAG index poisoning | Mitigated | Memory integrity via Merkle chains |
| Jailbreak attacks | Mitigated | Semantic firewall 27 pattern categories |