ClawHub
Back to skill

Security audit

Insurance Digital Live

Security checks for vulnerabilities and agentic risk

Overview

This is a disclosed insurance sales-support skill that provides draft scripts and compliance checklists, with no executable code or hidden access to user systems.

Install only if you need insurance live-streaming sales support for mainland China. Treat all generated scripts as drafts, avoid entering real customer identifiers or screenshots, verify current regulatory claims with official or institutional sources, and obtain compliance approval before publishing or using lead-capture follow-up flows.

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 (2)

Intent-Code Divergence

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The skill states it does not access live-platform or customer data, yet later encourages workflows that solicit age details and move prospects into private channels such as enterprise WeChat and groups. Even if framed as user-executed rather than skill-executed, this contradiction weakens safety boundaries and can normalize collection and off-platform handling of personal data without clear minimization, consent, or retention controls.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Low
Category
Not specified by scanner
Confidence
79% confidence
Finding

The execution-boundary table describes the code blocks as merely structured storage for reading and rewriting and says the skill does not run them. But the document uses these blocks as substantive operational logic for compliance checking, engagement prompts, and conversion workflows, so the documentation minimizes what they actually encode and how they are intended to drive behavior.

Content

No source excerpt is available for this finding.

Static analysis

No suspicious patterns detected.