Security audit

Docker Sandbox

Security checks across malware telemetry and agentic risk

Overview

This is an instruction-only Docker sandbox guide with powerful but disclosed, user-directed commands; users should be careful with mounted workspaces and reset commands.

Install only if you intend to use Docker Desktop sandbox features. When testing untrusted code, mount a disposable or carefully chosen workspace, use deny-by-default network rules where practical, avoid detached runs unless needed, and treat reset/rm as destructive cleanup operations.

SkillSpector

By NVIDIA

Vulnerability Patterns

Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep

Findings (2)

Missing User Warnings

Medium

Confidence: 94% confidence
Finding: The skill documents `docker sandbox reset` as a troubleshooting step without clearly warning that it clears all sandbox state, which can lead users to destroy environments and lose work unintentionally. In a tool explicitly marketed for experimentation and persistent dev environments, users may reasonably run this command during recovery and only discover the data loss afterward.

Missing User Warnings

Medium

Confidence: 97% confidence
Finding: The skill emphasizes VM isolation but explains that the host workspace is mounted into the sandbox without clearly warning that commands run inside the sandbox can read, modify, or delete those host files. This can create a false sense of safety when using the sandbox for untrusted code, because the mounted project remains directly exposed despite the VM boundary.

VirusTotal

66/66 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.