ClawHub

Clawscan Vigil

Security checks across malware telemetry and agentic risk

Overview

The skill is a local security scanner, but it needs Review because dynamic scanning executes scanned skill code by default inside the scanner process with weak containment.

Install only if you are comfortable scanning untrusted skills with dynamic analysis disabled or in a disposable environment. Prefer running with --no-dynamic unless you trust the target skill or can sandbox the scanner process; also expect local quota/license files under ~/.clawscan.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Behavioral ASTexec() Call, eval() Call, Dynamic Import
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (2)

exec() call detected

High
Category
Dangerous Code Execution
Content
return findings
            
            # Execute in sandbox
            exec(compiled, restricted_globals)
            self.trace.completed = True
            
        except Exception as e:
Confidence
94% confidence
Finding
exec(compiled, restricted_globals)

Tp4

High
Category
MCP Tool Poisoning
Confidence
77% confidence
Finding
The documented behavior goes beyond pure security scanning into licensing, quota enforcement, local state storage under the user home directory, and premium feature gating, none of which are clearly reflected in the declared purpose. This mismatch can mislead users about what the skill actually does and what data or files it may modify, reducing informed consent and trust.

VirusTotal

64/64 vendors flagged this skill as clean.

View on VirusTotal