ClawHub
Back to skill

Security audit

Whiteboard Prompt Crafter

Security checks for vulnerabilities and agentic risk

Overview

The skill itself is a simple whiteboard image-prompt helper, but its metadata advertises unrelated high-impact capabilities like wallet, purchases, crypto, and sensitive credentials.

Review before installing. Use it only for whiteboard-style image generation, expect image prompts and reference images to be processed by image-generation and messaging tools, and do not grant wallet, purchase, crypto, or sensitive-credential access unless the publisher provides a corrected, purpose-aligned explanation.

SkillSpector

Vulnerability Patterns
  • 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
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (3)

Description-Behavior Mismatch

Medium
Confidence
95% confidence
Finding
The prompt templates explicitly request environmental/background details such as walls, office blur, trays, lighting, and camera angle, which conflicts with the skill's stated behavior that the output should contain only the whiteboard and no background. This mismatch can cause the agent to generate outputs that violate user expectations or downstream product constraints, especially if callers rely on the manifest for safe, bounded image generation behavior.

Description-Behavior Mismatch

Medium
Confidence
96% confidence
Finding
Multiple templates continue to encode scene-setting instructions inconsistent with a whiteboard-only output, creating a systematic specification drift rather than a one-off wording issue. In an agent skill, this can lead to repeated policy or product-contract violations across many invocations, making the behavior unreliable and harder to control.

Vague Triggers

Medium
Confidence
85% confidence
Finding
The trigger phrases are broad enough to activate on generic image-generation requests such as '这种风格的图怎么做', which can cause the skill to intercept requests outside its narrowly intended whiteboard-prompt use case. This can lead to incorrect routing, unintended tool use, or the skill shaping user requests into a specific workflow without clear user intent.

Static analysis

No suspicious patterns detected.