ClawHub
Back to skill

Security audit

文件工具箱

Security checks across malware telemetry and agentic risk

Overview

This is a local file-organization skill with some broad and inconsistent documentation, but the inspected artifact does not show hidden, deceptive, or exfiltrating behavior.

Install only if you are comfortable giving the agent local file read/write and command execution authority for the folders you name. Use preview mode first, avoid applying the sample deletion setting without review, and do not rely on the local-only claim unless your agent platform itself keeps prompts and file metadata local.

SkillSpector

Vulnerability Patterns
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Behavioral ASTexec() Call, eval() Call, Dynamic Import
  • 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
Findings (4)

Intent-Code Divergence

Medium
Confidence
85% confidence
Finding
The skill presents conflicting privacy and execution claims: it advertises local-only operation while also accepting a callback URL and later describing network troubleshooting behavior. In a file-management skill with read/write/exec permissions, ambiguous statements about whether metadata or results may be sent off-host can mislead users and agents into exposing sensitive filesystem information.

Intent-Code Divergence

Medium
Confidence
92% confidence
Finding
The dependency section says the free version is fully local with no extra API key and no file uploads, yet also lists an LLM API as required. This contradiction can conceal real data exposure paths, especially because natural-language file analysis may send filenames, paths, or content-derived metadata to a remote model service.

Vague Triggers

Medium
Confidence
76% confidence
Finding
The capability coverage text is overly broad and includes vague trigger phrases such as file handling, document conversion, format conversion, and content extraction, even though the rest of the skill focuses on file organization. Loose activation criteria increase the chance an agent invokes this skill in inappropriate contexts, expanding access to read/write/exec operations beyond the intended scope.

Missing User Warnings

Medium
Confidence
89% confidence
Finding
The sample configuration includes automatic deletion of files older than 90 days without a prominent safety warning, confirmation requirement, or default dry-run safeguard. In a skill that manages user files and supports exec/write actions, this can lead to irreversible data loss if copied verbatim or invoked by an agent without careful review.

VirusTotal

64/64 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.