ClawHub
Back to skill

Security audit

github-pages-publish

Security checks across malware telemetry and agentic risk

Overview

The skill is purpose-aligned for publishing static sites, but it uses GitHub write access to create public pages and overstates that it will never overwrite existing remote content.

Install only if you intend to publish static files publicly on GitHub Pages. Confirm the exact directory, repository, branch, and remote path before running it, avoid directories containing secrets or private reports, and prefer a narrowly scoped fine-grained GitHub token with Contents read/write for only the target repository.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • 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
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (2)

Intent-Code Divergence

Medium
Confidence
97% confidence
Finding
The docstring states the directory upload 'can never wipe out pre-existing remote content,' but the implementation updates any remote file whose path matches a local file by supplying the existing SHA and issuing a PUT. That can overwrite important existing content such as index.html or other published assets, so the safety claim is misleading and may cause users to trust a destructive operation more than they should.

Vague Triggers

Medium
Confidence
92% confidence
Finding
The README advertises very broad, natural-language trigger phrases such as “把这个网页发到网上” and “生成一个能分享的网页链接”, which can cause the skill to be invoked in situations the user may not clearly understand as a public deployment action. Because this skill publishes content to GitHub Pages and creates a publicly accessible URL, accidental invocation can lead to unintended disclosure of private or internal files that the user only meant to preview or share privately.

VirusTotal

62/62 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.