ClawHub
Back to skill

Security audit

OpenClaw代码审查助手

Security checks across malware telemetry and agentic risk

Overview

This skill is non-destructive, but it claims to perform real code and security reviews while returning canned results unrelated to the user's code.

Install only if you treat this as a demo or placeholder. Do not rely on its reports for merge decisions, CI gates, or security assurance, and do not provide a GitHub token unless the implementation is replaced with real, transparent PR analysis and least-privilege handling.

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 (2)

Intent-Code Divergence

Medium
Confidence
98% confidence
Finding
The review command accepts a user-supplied target and presents itself as performing code review, but it never inspects that target and instead emits a hardcoded report. This is dangerous because users may rely on fabricated security findings and a false sense of review coverage, causing real vulnerabilities in the actual codebase to be missed before merge or deployment.

Intent-Code Divergence

Medium
Confidence
97% confidence
Finding
The diff command claims to analyze the current git diff, but only displays git diff statistics and then prints a fabricated issue summary unrelated to repository contents. In a code-review security skill, this can mislead developers into trusting nonexistent review results, which may allow insecure changes to pass unexamined.

VirusTotal

65/65 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.