ClawHub
Back to skill

Security audit

中文公文写作

Security checks across malware telemetry and agentic risk

Overview

This skill is a coherent Chinese official-document writing helper with optional local draft linting, and I found no hidden persistence, credential use, exfiltration, or destructive behavior.

Install this if you need help drafting or reviewing Chinese official/work documents. Be careful when providing confidential drafts, DOCX files, or government/business materials: the skill is designed to process user-supplied text and may use an optional local lint script on files you name, so only pass documents you intend the agent to inspect.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (3)

Lp3

Medium
Category
MCP Least Privilege
Confidence
88% confidence
Finding
The skill instructs the agent to read numerous local reference files and even mentions checking local .txt/.md/.docx drafts, which implies file-read capability despite no declared permissions. Undeclared file access is dangerous because it expands the skill’s effective privilege surface and can lead to unintended access to local data outside the user’s expectations or policy controls.

Tp4

High
Category
MCP Tool Poisoning
Confidence
96% confidence
Finding
The declared description says the skill drafts, rewrites, compresses, and reviews official Chinese documents, but the analyzed behavior indicates it mainly performs heuristic checking plus local file scanning/CLI-style input handling not disclosed in the description. This mismatch is security-relevant because users and orchestration systems may grant trust, route sensitive content, or skip safeguards based on the stated purpose while the actual behavior accesses files and fails to provide the promised controls.

Natural-Language Policy Violations

Medium
Confidence
90% confidence
Finding
The skill instructions are entirely in Chinese and define behavior for drafting and revising Chinese official documents without providing a user language-choice mechanism. This can override user preference or platform language expectations, causing unsafe or unusable outputs in multilingual settings and making it harder for users to understand safety-critical omissions or caveats.

VirusTotal

64/64 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.