guava-guard

You can scan AgentSkills for malware by running a static-analysis scanner on the skill directory to detect malicious code patterns, credential theft, and prompt injection. This prevents compromised skills from executing in your AI agent workspace.

When auditing third-party AI agent skills, look for prompt injection, malicious code execution, credential theft, and obfuscation techniques. Detecting these threat categories prevents malicious payloads from compromising your workspace security.

No, an AgentSkill security scanner with context-aware analysis does not flag documentation as malicious. It restricts pattern matching to executable files like js, py, and sh, reducing false positives by about 80%.

No, you do not need external dependencies to run a prompt injection and malware detection scan on agent skills. The scanner operates with zero external dependencies, using context-aware pattern matching to identify threats locally.

Yes, you can scan a ClawHub registry to prevent fake prerequisites and credential theft. Auditing skill directories verifies each third-party skill is safe, catching threats like fake prerequisites and credential-stealing malware before execution.