ClawHub
Back to skill

Security audit

gtrk 直播切片

Security checks for vulnerabilities and agentic risk

Overview

This video-editing skill needs review because conflicting retry instructions could cause additional charges without renewed approval.

Before use, require refund confirmation or your explicit approval before any paid resubmission. Recover existing tasks after timeouts. Confirm the quoted budget and allow cloud processing only for recordings you are comfortable sharing.

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)

The troubleshooting workflow directs an unconditional retry despite the documented refund-confirmation requirement.

Medium
Category
Excessive Agency
Confidence
96% confidence
Finding

The troubleshooting instructions require retrying a failed segment once without first checking the refund receipt, contradicting the billing section's explicit refund-confirmation requirement. Because a fresh submission can incur another charge, an agent following this more specific recovery instruction could spend additional user funds without authorization. The one-retry limit and segment-level isolation reduce the potential loss, but do not eliminate it.

Content

No source excerpt is available for this finding.