Security audit

网页抓取引擎

Security checks for vulnerabilities and agentic risk

Overview

This is a web scraping helper, but it includes under-scoped guidance for login tokens, proxy rotation, and stealth scraping that users should review carefully.

Install only if you intend to use an agent for web scraping and code execution. Before use, confirm you have authorization for each target site, avoid private or login-protected data unless explicitly permitted, do not use stealth/proxy rotation to bypass site restrictions, and review any generated scripts before allowing them to run, write files, upload data, or use API keys/cookies.

SkillSpector

Vulnerability Patterns

Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
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
Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access

Findings (4)

Intent-Code Divergence

High

Confidence: 98% confidence
Finding: The skill claims it is limited to publicly accessible content, but elsewhere it instructs users to inject cookies/tokens, maintain authenticated sessions, use proxy rotation, and stealth techniques to bypass anti-bot protections. This creates a clear mismatch between stated safety boundaries and documented behavior, increasing the risk of unauthorized access, policy evasion, and misuse for scraping protected resources.

Intent-Code Divergence

Medium

Confidence: 91% confidence
Finding: The skill states that only whitelisted commands should be executed, but the skill itself exposes exec capability and provides runnable scripts without any actual command allowlist, validation layer, or enforcement mechanism. This can mislead operators into assuming safety controls exist when in practice arbitrary code or shell execution may be enabled through the agent environment.

Vague Triggers

Medium

Confidence: 83% confidence
Finding: The trigger keywords are broad and cover generic scraping and automation terms, which raises the likelihood that the skill is invoked in contexts the user did not explicitly intend. Because the skill has exec/write capabilities and supports network scraping workflows, accidental activation can lead to unintended code generation, file creation, or higher-risk collection behavior.

Missing User Warnings

Medium

Confidence: 90% confidence
Finding: The skill supports file creation, database writes, uploads, and execution, but the description does not prominently warn users about system or data impact before those actions are suggested or performed. In an agent context, this can normalize high-impact operations without informed consent, increasing the chance of accidental persistence, data exfiltration, or modification of local resources.

Static analysis

No suspicious patterns detected.