ClawHub
Back to skill

Security audit

安全扫描器(免费版)

Security checks for vulnerabilities and agentic risk

Overview

This is a legitimate security-scanning skill, but its sample execution path is unsafe because user-supplied targets are interpolated into shell commands with exec access.

Install only if you will use it for assets you are authorized to test. Before running any generated script or command, validate targets strictly and replace shell=True/raw command strings with structured subprocess argument lists. Expect active network traffic and local report files containing potentially sensitive scan results.

SkillSpector

Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Output HandlingUnvalidated Output Injection, Cross-Context Output, Unbounded Output
  • Tool MisuseTool Parameter Abuse, Chaining Abuse, Unsafe Defaults
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Findings (5)

Intent-Code Divergence

Medium
Confidence
73% confidence
Finding
The documentation says scanning runs locally and needs no API key, but the skill clearly performs active network interactions against user-supplied targets using nmap, nuclei, nikto, and sslscan. This can mislead users and downstream agents into treating the skill as low-risk/local-only when it actually initiates external network activity that may violate authorization boundaries or policy.

Vague Triggers

Medium
Confidence
78% confidence
Finding
The trigger conditions are broad enough to invoke the skill for generic 'security detection' or 'compliance audit' requests without clearly constraining scope, authorization, or target ownership. In an agent setting, ambiguous triggers increase the chance of unintended active scanning against real systems based on vague user prompts.

Missing User Warnings

Medium
Confidence
88% confidence
Finding
The skill description does not prominently warn that it will perform active network scans and write reports/results to local disk. In an agent environment with exec access, hidden side effects materially increase operational risk because users may believe they are requesting passive analysis rather than network actions and persistent artifact creation.

Unvalidated Output Injection

High
Category
Output Handling
Content
def _run_command(self, command):
        """执行系统命令"""
        try:
            result = subprocess.run(
                command, shell=True, capture_output=True,
                text=True, timeout=600
            )
Confidence
99% confidence
Finding
The skill builds shell commands with interpolated target data and executes them via subprocess.run(..., shell=True). If an attacker controls or influences the target/input, they can inject shell metacharacters to execute arbitrary commands on the host running the agent, making this especially dangerous because the skill exposes exec capability and is designed to accept user-provided targets.

Tool Parameter Abuse

High
Category
Tool Misuse
Content
def _run_command(self, command):
        """执行系统命令"""
        try:
            result = subprocess.run(
                command, shell=True, capture_output=True,
                text=True, timeout=600
            )
Confidence
98% confidence
Finding
The combination of exec capability and shell-based command construction allows parameter abuse beyond the intended scan options. A malicious prompt or crafted target could alter command semantics, chain additional commands, redirect output, or invoke unrelated tools, leading to arbitrary execution or unauthorized scanning behavior.

Static analysis

No suspicious patterns detected.