ClawHub
Back to skill

Security audit

Reference Formatter

Security checks across malware telemetry and agentic risk

Overview

This skill formats bibliography text into citation files, and its file-writing and optional lookup behavior fit that purpose.

Use explicit output filenames or locations, especially in shared workspaces. Keep private drafts, internal URLs, annotations, or unpublished bibliography data offline unless you intentionally approve web lookup for missing citation details.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (3)

Intent-Code Divergence

Medium
Confidence
82% confidence
Finding
Step 4 instructs the skill to generate output files immediately without user confirmation, despite the skill metadata indicating confirmation should occur first. In an agent environment with file-write capabilities, this can cause unintended persistence of user data to disk and surprise file creation, especially when users expected a preview-only transformation step.

Missing User Warnings

Medium
Confidence
92% confidence
Finding
The skill defaults to writing files into the workspace root without clearly warning the user or obtaining consent for the write target. In agentic contexts, silent file creation can overwrite expectations, persist sensitive bibliography content locally, and interact badly with automated workflows that monitor the workspace.

Missing User Warnings

Medium
Confidence
95% confidence
Finding
The skill permits web_search to complete missing citation fields without warning that user-supplied reference data may be sent to external services. Bibliographies can contain unpublished manuscripts, internal URLs, private annotations, or sensitive research topics, so silent network transmission creates a real confidentiality and privacy risk.

VirusTotal

64/64 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.