ClawHub
Back to skill

Security audit

Sendbird

Security checks for vulnerabilities and agentic risk

Overview

This is a disclosed staff moderation helper for ClawHub admin actions, with clear confirmation, reason, auth, audit, and verification expectations.

Install only for authorized ClawHub staff moderation. Confirm the exact target and reason before any write command, and be especially careful with ban, unban, role-change, and hide/restore actions because they affect real users and public content.

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)

Intent-Code Divergence

Medium
Confidence
96% confidence
Finding
The safety guidance says untagged actions are read-only, but `unban_user` is listed without a `[write]` tag even though it changes moderation state. This can cause an agent to execute a state-changing action without confirmation, leading to unauthorized reversal of access controls in a live messaging environment.

Static analysis

No suspicious patterns detected.