ClawHub
Back to skill

Security audit

Smart Files

Security checks across malware telemetry and agentic risk

Overview

This skill is a local file-management helper with clearly disclosed privacy and destructive-operation risks when used outside dry-run mode.

Install only if you want an agent-accessible local file search and organization tool. Do not run content search on directories containing secrets or personal data, and treat --watch --force as destructive because it moves files and deletes originals without rollback. Prefer dry-run first and pass narrowly scoped directories.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • 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
Findings (4)

Lp3

Medium
Category
MCP Least Privilege
Confidence
84% confidence
Finding
The skill advertises broad filesystem features and explicitly documents use of the SMART_FILES_WORKSPACE environment variable, but the metadata shown does not declare corresponding permissions. In an agent setting, undeclared environment access weakens operator visibility and consent, and can expose sensitive path or configuration data unexpectedly when the skill is invoked.

Tp4

High
Category
MCP Tool Poisoning
Confidence
95% confidence
Finding
The declared description frames the skill mainly as a file-management/search utility, but the documented behavior includes continuous directory watching, persistent journaling, and destructive mutation with copy-then-unlink deletion under --force. In an autonomous agent context, this mismatch is dangerous because users or orchestrators may grant or invoke the skill expecting read-oriented analysis while it can actually move, rename, and delete files, increasing the risk of unintended data loss or abuse.

Description-Behavior Mismatch

Medium
Confidence
95% confidence
Finding
The watch mode can perform destructive mutations by copying files into category folders and then unlinking the originals when --force is used. While the code warns on the CLI, the skill description emphasizes search/management and does not clearly surface that continuous watch mode can move and effectively delete originals, which increases the risk of unsafe invocation by an agent or user expecting read-only behavior.

Context-Inappropriate Capability

Medium
Confidence
98% confidence
Finding
The path-resolution logic only warns when a path is outside the detected workspace but still proceeds, allowing scanning and file operations against arbitrary directories. In an agent skill context, this weak boundary check is dangerous because agent-supplied arguments could cause inspection or modification of sensitive host files outside the intended project scope.

VirusTotal

64/64 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.