ClawHub
Back to skill

Security audit

AI KEY·有什么到期了

Security checks for vulnerabilities and agentic risk

Overview

This reminder skill uses disclosed local storage and explicitly leaves actions to the user, although some workflow wording is inconsistent.

Use this as a reminder tool: confirming a condition should not authorize publishing or business changes. Verify the missing configuration and reference documents, check the storage folders, and keep sensitive details out of persistent notes. Clarifying the repeated action-execution wording would improve reliability.

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 due-processing workflow directs action execution despite an explicit reminder-only prohibition.

Medium
Category
Excessive Agency
Confidence
98% confidence
Finding

The due-processing workflow repeatedly instructs the agent to execute a stored action once its condition is satisfied, conflicting with the explicit reminder-only prohibition against executing actions for the user. A tool-enabled agent could interpret confirmation of the condition as authorization to perform the action, although the later prohibition reduces this risk and there is no evidence of malicious intent.

Content

No source excerpt is available for this finding.