ClawHub
Back to skill

Security audit

图解

Security checks for vulnerabilities and agentic risk

Overview

This diagram skill is not clearly malicious, but it asks for broad read/write/exec authority while its instructions are inconsistent and under-scoped.

Review this skill before installing. It appears to be a diagram helper rather than malware, but its requested read/write/exec authority is broader than its documented behavior explains. Use it only in a sandboxed or low-trust workspace unless the publisher narrows activation to explicit diagram requests, fixes the output schema, and documents any command execution clearly.

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 documented output schema describes a grading/evaluation report with fields like overall_grade, total_score, and safety comments, which is inconsistent with a diagram-generation skill. This mismatch can mislead an agent into producing or trusting the wrong output type, causing downstream automation errors, incorrect tool chaining, or accidental processing of unrelated evaluation data instead of diagram content.

Intent-Code Divergence

Medium
Confidence
89% confidence
Finding
The FAQ states that no extra software is needed, while the manifest grants exec capability and the document discusses command execution and environment dependencies. This inconsistency can cause operators and calling agents to underestimate the execution surface, leading to unsafe deployment decisions, insufficient sandboxing, or unexpected command execution in environments assumed to be non-executing.

Vague Triggers

Medium
Confidence
84% confidence
Finding
The skill description uses broad project-management and collaboration language that overlaps with many ordinary conversations, increasing the chance the agent invokes this skill when the user did not actually request diagram generation. Over-broad triggering is dangerous because the skill has read/exec/write capabilities, so accidental activation expands the attack surface and may cause unnecessary file access or command execution in unrelated contexts.

Static analysis

No suspicious patterns detected.