ClawHub
Back to skill

Security audit

chat-ai

Security checks across malware telemetry and agentic risk

Overview

This text-to-SQL skill is not clearly malicious, but it delegates database query execution to unbundled local code with unclear approval, credential, and read-only boundaries.

Install only if you can inspect and trust the external Winner-Ai project this wrapper imports. Use a read-only, least-privilege database account, require users to review generated SQL before execution, and avoid sending sensitive business data unless retention and output handling are clear.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
Findings (1)

Missing User Warnings

Medium
Confidence
96% confidence
Finding
The wrapper defines a user_input_callback that automatically returns '同意' and then invokes a generic external orchestrator for natural-language-to-SQL execution. If the underlying orchestrator uses user confirmation as a safety gate for risky SQL generation/execution or privileged actions, this code silently bypasses that control and can permit unintended queries or state-changing operations without real user consent.

VirusTotal

65/65 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.