Back to skill

Security audit

SEO审计大师

Security checks across malware telemetry and agentic risk

Overview

This is a legitimate SEO audit guide, but it needs review because some examples handle API tokens and outbound submissions unsafely.

Install only if you are comfortable reviewing the examples before use. Treat all webmaster tokens as secrets, store them outside source files, prefer HTTPS endpoints, and avoid enabling automatic URL submission hooks unless you explicitly want every new post URL sent to that third party.

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 (4)

Intent-Code Divergence

Medium
Confidence
95% confidence
Finding
The skill presents contradictory security guidance: one section says no extra API key configuration is required, while later sections instruct users to configure API keys or tokens. This can mislead operators into unsafe setup practices, weaken review quality, and cause accidental exposure or misuse of credentials during deployment.

Intent-Code Divergence

High
Confidence
99% confidence
Finding
The document claims tokens should only be stored in environment variables and never written to outputs, but its examples include hardcoded token placeholders in executable code. This normalizes insecure patterns and increases the likelihood that users will embed real secrets in source files, logs, or shared snippets.

Intent-Code Divergence

High
Confidence
97% confidence
Finding
The skill's security guidance says only whitelisted commands should be executed and user input should not be concatenated into shell commands, yet multiple examples interpolate domains, URLs, or other user-controlled values into curl, sed, and similar commands. In an agent environment with exec capability, this can directly enable command injection or unsafe outbound requests if an implementation follows the examples naively.

External Transmission

Medium
Category
Data Exfiltration
Content
add_action("publish_post", "baidu_push");
function baidu_push($post_id) {
    $url = get_permalink($post_id);
    $api = "http://data.zz.baidu.com/urls?site=blog.example.com&token=YOUR_TOKEN";
    $ch = curl_init();
    curl_setopt($ch, CURLOPT_URL, $api);
    curl_setopt($ch, CURLOPT_POST, true);
Confidence
92% confidence
Finding
http://data.zz.baidu.com/

VirusTotal

VirusTotal findings are pending for this skill version.

View on VirusTotal

Static analysis

No suspicious patterns detected.