ClawHub
Back to skill

Security audit

gtrk 智能口播剪辑

Security checks for vulnerabilities and agentic risk

Overview

The skill performs video editing but also starts additional production work by default, even when the user requested only a basic edit.

Explicitly limit the task to the requested editing project and require approval before storyboard or further production steps. Expect extracted audio, supplied script content, and optionally a 720p video proxy to reach the cloud service, with cloud usage potentially consuming account credit. Use trusted dependency downloads and independently sourced checksums.

A.I.G

Vulnerability Patterns
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs

SkillSpector

Vulnerability Patterns
  • 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
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (1)

Automatically expands project generation into downstream storyboard and production workflows without an affirmative user request.

Medium
Category
Excessive Agency
Confidence
97% confidence
Finding

The skill explicitly directs the agent to continue from the requested editing project into storyboard dispatch and downstream production without asking, treating the absence of an opt-out as authorization. This crosses the user's task boundary and could cause unintended tool executions, additional artifacts, or downstream resource consumption; the explicit stop conditions reduce risk but do not establish affirmative consent. Automatic continuation is reasonable when the user requested the full production workflow, but this skill also targets narrower requests such as removing pauses or generating an editable project.

Content

No source excerpt is available for this finding.

Static analysis

No suspicious patterns detected.