ClawHub
Back to skill

Security audit

clawvision

Security checks across malware telemetry and agentic risk

Overview

ClawVision is a local session-summary exporter whose sensitive session access is disclosed and tied to explicit user-controlled export requests.

Install only if you are comfortable with a user-invoked skill reading the chosen OpenClaw session and writing local summary/export files. Avoid running it on sessions containing secrets, credentials, personal data, or internal identifiers unless you have reviewed the scope and are comfortable exporting a sanitized summary.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • 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
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (1)

Tp4

High
Category
MCP Tool Poisoning
Confidence
92% confidence
Finding
The top-level description frames the update as a harmless transparency change, but the skill actually has broad operational behavior: reading session history, invoking local inference, executing a script, and writing multiple export artifacts. This mismatch can mislead reviewers and users about the true data-handling and execution surface, increasing the risk of sensitive conversation data being exported or processed without fully informed consent.

VirusTotal

65/65 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.