ClawHub
Back to skill

Security audit

LYGO Protocol Stack Operator

Security checks across malware telemetry and agentic risk

Overview

This is a documented LYGO stack helper with local verification scripts and visible optional operational commands, not hidden or automatic harmful behavior.

Install this only if you intend to work with the LYGO stack. Run the local byte gate on files you choose, set LYGO_STACK_ROOT only to a trusted checkout, and treat clone, pip install, Docker, node API, anchor worker, HF upload, git push, and ClawHub publish commands as explicit external actions that need your approval.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • 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 (3)

Lp3

Medium
Category
MCP Least Privilege
Confidence
93% confidence
Finding
The skill advertises and instructs use of shell commands, file access, environment-variable usage, Docker, git, pip, and local server execution, yet declares no permissions. This creates a transparency and consent problem: an agent or reviewer may underestimate the skill's ability to read local files, access env-derived secrets, or execute system commands, increasing the chance of unintended privileged actions.

Tp4

High
Category
MCP Tool Poisoning
Confidence
90% confidence
Finding
The skill makes broad claims about audits, node APIs, mesh security, attestation, TLS public mesh, and safe orchestration, but the provided content is primarily documentation and command suggestions rather than an implemented, verifiable control framework. This overstatement can mislead users into trusting nonexistent security guarantees or operational safeguards, causing them to run commands, ingest data, or expose services under false assumptions of protection.

Intent-Code Divergence

Medium
Confidence
95% confidence
Finding
The quick-reference includes direct publish and push commands for Hugging Face and ClawHub without any documented approval checkpoint, which conflicts with the stated safety property that publish/post actions require human approval. In an agent skill context, concise operational docs are often treated as executable guidance, so these commands can normalize or enable unattended release actions if the surrounding workflow does not enforce an external gate.

VirusTotal

65/65 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.