ClawHub
Back to skill

Security audit

My Awesome Tools

Security checks across malware telemetry and agentic risk

Overview

The tools are mostly simple local generators, but the code review assistant appears to present canned results as if it reviewed real code.

Install only if you are comfortable reading and running local shell scripts. Use the generators in a scratch or backed-up directory, do not rely on the code-review assistant for real security or pull request review, and review any generated HTML before publishing it publicly.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (3)

Intent-Code Divergence

Medium
Confidence
98% confidence
Finding
The script presents itself as a code review and PR analysis assistant, but the `review` and `diff` paths mostly emit static, hardcoded reports instead of analyzing the provided target. This can mislead users into believing security review was performed when no real inspection occurred, creating a false sense of assurance and potentially allowing vulnerable code to pass unchecked.

Missing User Warnings

Medium
Confidence
97% confidence
Finding
The script unconditionally writes generated content to README.md using shell redirection, which will overwrite any existing file without prompting, backup, or safety check. In an agent or automation context, this can cause accidental loss of repository documentation or clobber a user-maintained README, especially when run in the wrong directory or with untrusted inputs.

Missing User Warnings

Low
Confidence
85% confidence
Finding
The generated page includes third-party scripts from public CDNs (Tailwind CDN, React, ReactDOM, Babel) without pinning integrity or warning the user that runtime behavior depends on remote code. If a CDN resource is compromised, changed unexpectedly, or blocked, users of the generated site could be exposed to client-side code execution, privacy leakage, or availability issues.

VirusTotal

65/65 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.