A runtime behavioral sandbox that detects guidance injection attacks in OpenClaw skills by observing what agents actually do instead of scanning what skills say
Existing OpenClaw skill scanners use static analysis and LLM-based content scanning to flag malicious skills before installation. The Trojan's Whisper paper (March 2026) proved that 94% of guidance injection attacks evade both approaches because the malicious payload is disguised as routine operational guidance, not explicit instructions. Meanwhile 12% of ClawHub's skill registry has been compromised at some point in 2026. The gap is clear. Instead of scanning skill text, this product spins up an isolated OpenClaw instance, installs the skill, runs a battery of natural user prompts, and observes what the agent actually does. Credential access, file writes outside sandbox, network exfiltration, privilege escalation attempts all get flagged as behavioral anomalies regardless of how the skill's guidance file describes them.
Demand Breakdown
Social Proof 4 sources
Gap Assessment
4 tools exist (VirusTotal Integration (built-in), Cisco DefenseClaw, SkillFortify, SecureClaw) but gaps remain: Only catches known malware signatures. Completely blind to guidance injection which uses natural language, not malware binaries. 94% evasion rate proven by Trojan's Whisper.; Uses static and LLM-based scanning. Falls into the exact category that Trojan's Whisper proved 94% evasion against. No runtime behavioral analysis..
Features3 agent-ready prompts
Competitive LandscapeFREE
| Product | Does | Missing |
|---|---|---|
| VirusTotal Integration (built-in) | Scans skill files against VirusTotal malware database on ClawHub upload | Only catches known malware signatures. Completely blind to guidance injection which uses natural language, not malware binaries. 94% evasion rate proven by Trojan's Whisper. |
| Cisco DefenseClaw | Open-source agent security governance framework with static skill scanning | Uses static and LLM-based scanning. Falls into the exact category that Trojan's Whisper proved 94% evasion against. No runtime behavioral analysis. |
| SkillFortify | Formal verification for AI agent skills using mathematical proofs of behavior | Formal verification works on well-defined properties but cannot model the emergent behaviors that guidance injection produces through context manipulation across prompts. |
| SecureClaw | Maps OpenClaw security posture to OWASP Agent Security Initiative standards | Compliance mapping tool, not a detection tool. Tells you what risks exist but does not actively block malicious skills at install time. |
Sign in to unlock full access.