ClawHub
Back to skill

Security audit

book-toolbook-html

Security checks for vulnerabilities and agentic risk

Overview

This skill is a disclosed HTML card-book generator, with limited caveats around its default remote cover image and placeholder promotional footer.

Before installing, know that generated pages will load a default Unsplash cover image unless you replace or remove it, which can disclose normal web request metadata to that image host. Also review or delete the footer promotion block if you do not want公众号/keyword call-to-action text in the final HTML.

SkillSpector

Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • 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 (3)

Description-Behavior Mismatch

Medium
Confidence
93% confidence
Finding
The template embeds a hardcoded remote Unsplash image in CSS, which creates an undeclared network dependency and causes client IP, user agent, and referrer-related metadata to be disclosed to a third party when the HTML is opened online. In the context of a 'single-file/offline/no external dependency' artifact, this is a real security and privacy issue because users may trust the file to be self-contained and open it in sensitive environments.

Context-Inappropriate Capability

Medium
Confidence
84% confidence
Finding
The template includes built-in公众号/keyword lead-generation content unrelated to the stated formatting function, which can silently inject promotional or traffic-diversion material into generated outputs. While not code execution, this is a trust and supply-chain integrity issue: users invoking a book-formatting skill may unknowingly publish operator-controlled marketing content in their documents.

Missing User Warnings

Low
Confidence
91% confidence
Finding
Loading a remote image without user-facing disclosure causes silent network access when the HTML is viewed, which can surprise users and leak basic access metadata to the remote host. This is especially problematic here because the skill advertises offline-friendly single-file output, so the undisclosed fetch conflicts with expected privacy properties.

Static analysis

No suspicious patterns detected.