ClawHub
Back to skill

Security audit

SEO blog

Security checks for vulnerabilities and agentic risk

Overview

This skill mostly generates blog files, but it also contains under-scoped instructions for writing and saving content inside a live CMS editor.

Review this before installing if you only want local blog asset generation. The UEESHOP editor automation section should be removed or split into a separate publishing skill with explicit user authorization and save/publish controls.

SkillSpector

Vulnerability Patterns
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Findings (4)

Intent-Code Divergence

Medium
Confidence
94% confidence
Finding
The skill claims in its completion section that it should only generate files and must not access UEESHOP, yet it later embeds detailed instructions and code for injecting content into a live TinyMCE editor, saving drafts, and verifying persistence. This contradiction can enable unintended browser automation against an external CMS and makes it easier for an agent to cross the boundary from offline content generation into live system manipulation.

Context-Inappropriate Capability

Medium
Confidence
97% confidence
Finding
The skill's stated purpose is to generate blog content files, but it includes operational browser-side code and procedural guidance for writing directly into the UEESHOP editor and persisting drafts. That expands the skill from content creation into CMS automation, creating a capability for unauthorized content modification or accidental publication workflow interference if an agent follows the embedded instructions.

Vague Triggers

Medium
Confidence
82% confidence
Finding
The skill metadata uses a broad activation/description pattern that can cause the agent to invoke this skill in loosely related contexts without strong trigger constraints. While this is not directly exploitative code execution, it increases the chance of unintended use, inappropriate data flow into the skill, or generation of content when the user did not clearly request this specific operation.

Natural-Language Policy Violations

Medium
Confidence
88% confidence
Finding
The default prompt hard-codes English output, which can override or conflict with the user's language preference if the orchestrator invokes the skill automatically. This can lead to policy misalignment, poor user experience, and unintended handling of multilingual requests, especially in systems expected to preserve user-specified locale or accessibility needs.

Static analysis

No suspicious patterns detected.