ClawHub
Back to skill

Security audit

图纸解析(专业版)

Security checks for vulnerabilities and agentic risk

Overview

The skill mostly matches its drawing-analysis purpose, but it asks for broad file, command, and webhook capabilities without clear enough scope or privacy controls for sensitive project drawings.

Review this skill carefully before installing. Use it only on drawing-analysis tasks, keep processing local unless you explicitly configure a trusted callback or webhook endpoint, and confirm dependencies and OCR tools before allowing command execution or package installation.

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

Medium
Confidence
97% confidence
Finding
The FAQ states that no extra packages are required, but the skill elsewhere instructs installing and using multiple external libraries such as ezdxf, pytesseract, pdf2image, and openpyxl. This discrepancy can mislead operators into enabling exec-based installation or running the skill in an unprepared environment, increasing the chance of unsafe ad hoc dependency installation and execution.

Vague Triggers

Medium
Confidence
84% confidence
Finding
The invocation guidance says to use the skill for project management, task planning, progress tracking, and team collaboration, which is much broader than drawing analysis. Overbroad triggers can cause the agent to activate this skill in unrelated contexts, unnecessarily granting read/exec/write-capable behavior and increasing the chance of unsafe file handling or command execution.

Missing User Warnings

Medium
Confidence
95% confidence
Finding
The skill supports callback_url and webhook outputs but does not clearly warn that extracted drawing content and metadata may be sent to external endpoints. In a document-processing context, that can expose proprietary building plans, annotations, or project details to third-party systems without informed consent.

Missing User Warnings

Medium
Confidence
94% confidence
Finding
The description advertises webhook-based automation but does not present a clear warning about automatic external transmission or downstream automation effects. Because the skill processes sensitive project artifacts, silent or poorly explained automation can propagate confidential data or trigger unintended actions in integrated systems.

Static analysis

No suspicious patterns detected.