ClawHub
Back to skill

Security audit

京东配置价格表生成器

Security checks for vulnerabilities and agentic risk

Overview

The skill’s main workflow is relevant, but its included script can create a misleading fixed spreadsheet and writes it to a developer-specific path instead of the user’s workspace.

Before installing or using this skill, review or modify the helper script so it uses the user’s actual hardware list, refreshed price sources, and a user-approved workspace output path. Treat generated pricing as a reference only and verify JD pages before purchasing.

SkillSpector

Vulnerability Patterns
  • 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
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (3)

Tp4

High
Category
MCP Tool Poisoning
Confidence
91% confidence
Finding
The skill claims to parse user-supplied hardware lists, search JD for current prices, and generate a corresponding Excel file, but the analyzed behavior is reportedly hardcoded and not driven by user input. This creates a trust and integrity problem: users may rely on fabricated or stale pricing/link data for purchasing decisions, and the fixed local output path also increases the chance of overwriting files or producing misleading artifacts unrelated to the requested configuration.

Description-Behavior Mismatch

Low
Confidence
90% confidence
Finding
The code writes output to a developer-specific absolute path, which can expose local filesystem structure, fail unpredictably on other systems, or overwrite files in an unintended location if that path exists. In an agent context, hard-coded local writes are risky because they bypass user choice and reduce transparency about where data is stored.

Missing User Warnings

Low
Confidence
84% confidence
Finding
The script saves an Excel file directly to disk without any prompt, confirmation, or configurable destination. While the file contents are not overtly dangerous, silent file creation can surprise users, create clutter, and in broader agent environments may violate expectations around consent and filesystem modification.

Static analysis

No suspicious patterns detected.