ClawHub
Back to skill

Security audit

Code Quality Tool Free

Security checks across malware telemetry and agentic risk

Overview

The skill is mostly a normal code-quality checker, but it under-discloses how much project code may be read or sent through an LLM and includes persistent Git hook setup.

Use this skill only on repositories you are comfortable exposing to your agent/LLM provider. Specify the target path and exclusions yourself, avoid scanning secrets/vendor/build directories, and do not run the pre-commit hook or patch-generation examples unless you have reviewed the exact changes first.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • 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)

Intent-Code Divergence

Medium
Confidence
78% confidence
Finding
The skill claims code will not be uploaded to external services while also declaring an LLM API as required. In an agent setting, that can mislead users into providing sensitive code under a false assumption of local-only processing, creating confidentiality risk if prompts or file contents are sent to a hosted model.

Vague Triggers

Medium
Confidence
85% confidence
Finding
The trigger phrase asks the agent to check the 'current project' without defining path, file-type, or permission boundaries. In a tool-enabled agent with exec and read access, that broad invocation can cause unintended large-scope scanning of sensitive repositories or unrelated files, increasing exposure and accidental over-collection.

Missing User Warnings

Medium
Confidence
88% confidence
Finding
The skill states that the agent may automatically generate repair patches but provides no warning, approval step, or limits around modifying user code. In practice, this can lead to unintended source changes, unsafe edits to security-sensitive files, or silent workflow disruption if users interpret the skill as read-only analysis.

VirusTotal

62/62 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.