ClawHub

openclaw-collab

Security checks across malware telemetry and agentic risk

Overview

This is a coherent collaboration helper, but it gives another agent broad persistent memory access and delegated publishing authority without enough scoping or approval guidance.

Install only if you intentionally want Hermes and OpenClaw to share persistent workspace state and delegate work. Inspect the referenced local scripts first, restrict writes to a dedicated collaboration folder, and require explicit human approval before publishing to ClawHub or sending tasks that can change accounts, files, or public content.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
Findings (2)

Missing User Warnings

Medium
Confidence
87% confidence
Finding
The skill explicitly recommends direct writes into a shared OpenClaw workspace and collaboration memory without guardrails about authorization, integrity, or side effects. In a multi-agent/shared-state environment, this can cause unintended task triggering, memory poisoning, or corruption of another agent's working context.

Missing User Warnings

Medium
Confidence
91% confidence
Finding
The publishing workflow instructs sending a local skill to an external service (clawhub.ai) and includes commands that may disclose code, metadata, and potentially embedded secrets, but gives no privacy, review, or approval warning. This creates a real exfiltration risk if sensitive files, proprietary logic, or credentials are present in the skill directory.

VirusTotal

63/63 vendors flagged this skill as clean.

View on VirusTotal