ClawHub
Back to skill

Security audit

LYRA OpenClaw Hybrid

Security checks across malware telemetry and agentic risk

Overview

This skill is mainly a public documentation and install map with consent-gated automation guidance, not an implementation that secretly accesses or changes user data.

Install only if you understand this is a high-level orchestration map for a broader LYRA/OpenClaw stack. Keep credentials in your own vault or environment, review any separately installed companion skills before use, and require explicit approval before browser writes, social posts, token launches, or memory writes.

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
79% confidence
Finding
The skill advertises a broad automation/orchestration role, but the analysis indicates additional local repository validation and secret-pattern scanning behavior not clearly disclosed in the description. Undisclosed local scanning can surprise operators, expand the trust boundary, and expose sensitive files or metadata during execution, especially in a skill explicitly tied to local secrets and operator machines.

VirusTotal

65/65 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.