ClawHub
Back to skill

Security audit

Write Literature Review

Security checks for vulnerabilities and agentic risk

Overview

This skill coherently helps users turn supplied literature files into an evidence-tracked review and Word document, with document parsing and conversion behavior disclosed in its instructions.

Before installing, users should be aware that this skill may convert and extract text from local documents they provide, and directory inputs may include many files recursively. Use a dedicated literature folder and enable OCR or external converters only for documents you intend to process.

SkillSpector

Vulnerability Patterns
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • 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)

Lp3

Medium
Category
MCP Least Privilege
Confidence
89% confidence
Finding
The skill instructs the agent to read and write files and execute shell commands, but it declares no corresponding permissions or capability boundaries. This creates a hidden execution surface where document parsing and DOCX generation may access local files or invoke external tools without transparent consent or sandboxing expectations, increasing the risk of unintended file exposure or command abuse.

Tp4

High
Category
MCP Tool Poisoning
Confidence
94% confidence
Finding
The documented behavior materially exceeds the high-level description by invoking external converters/OCR tools, processing additional file formats, exporting extracted text and manifests, and generating output artifacts beyond what a user would reasonably infer from the description. This mismatch undermines informed consent and can hide risky processing paths such as shelling out to caj2pdf, tesseract, pdftotext, or LibreOffice, which expand the attack surface and may process sensitive local content unexpectedly.

Static analysis

No suspicious patterns detected.