ClawHub
Back to skill

Security audit

艺术

Security checks across malware telemetry and agentic risk

Overview

This skill is mainly for a shared pixel art board, but it asks for broad command/file authority and persistent credentials with unclear scoping.

Review this before installing. Only use it in a sandbox or trusted workspace, confirm what script will actually be executed, avoid putting secrets or personal data in chat or state files, and remove or protect ~/.config/artboard/credentials.json if you stop using the skill.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Findings (6)

Intent-Code Divergence

Medium
Confidence
95% confidence
Finding
The skill contradicts itself by classifying usage as pure Markdown/natural-language driven while elsewhere requiring shell execution and exec-capable tooling. This misrepresentation can cause an agent or operator to grant the skill a safer trust level than warranted, increasing the chance that command execution occurs without appropriate review or sandboxing.

Intent-Code Divergence

Medium
Confidence
93% confidence
Finding
The document claims command execution is restricted to a whitelist, but the declared tooling and examples expose generic exec/bash behavior with no enforceable whitelist shown. This creates a false sense of safety around arbitrary command execution, which could lead to abuse if user-controlled input reaches shell commands.

Vague Triggers

High
Confidence
90% confidence
Finding
The description contains an overly broad invocation cue about general AI/model orchestration use that is unrelated to pixel-canvas functionality. Overbroad routing language can cause the skill to be invoked in inappropriate contexts, unnecessarily exposing exec, file-write, and credential-handling behaviors outside its intended domain.

Missing User Warnings

Medium
Confidence
91% confidence
Finding
The skill documents that registration stores a bot ID and authentication token in a local credentials file, but it does not present a clear, upfront warning about persistent credential storage. In agent environments, silent persistence of tokens can increase the risk of accidental disclosure, reuse by other tools, or unsafe filesystem permissions.

Missing User Warnings

Medium
Confidence
88% confidence
Finding
The skill instructs the agent to maintain persistent project state in a local JSON file without clearly warning users that this data survives across sessions. Persistent state can unintentionally retain sensitive prompts, activity history, or operational details that other processes or users may later read.

Credential Access

High
Category
Privilege Escalation
Content
## 功能能力
### 1. 机器人注册与凭证管理
通过 `artboard.sh register "YourBotName" "What kind of art you make"` 注册机器人，
凭证自动保存到 `~/.config/artboard/credentials.json`。注册后通过 `artboard.sh test`
验证 API 连接正常。凭证文件包含 bot ID 和认证 token，用于后续所有 API 操作.

### 2. 像素放置与冷却管理
Confidence
94% confidence
Finding
credentials.json

VirusTotal

VirusTotal findings are pending for this skill version.

View on VirusTotal

Static analysis

No suspicious patterns detected.