ClawHub
Back to skill

Security audit

构建 AI Agent 实战指南

Security checks across malware telemetry and agentic risk

Overview

This is a documentation-only skill for designing AI agents and does not add code execution, credentials access, persistence, or hidden data flows.

Before installing, expect this skill to guide agent architecture decisions rather than implement an agent. Review its advice against your own security, authorization, and human-review requirements for any real system that can take actions.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • 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)

Vague Triggers

Medium
Confidence
89% confidence
Finding
The skill advertises many broad trigger phrases such as '帮我设计一个 agent' and related variants without clear exclusions, which can cause over-activation in loosely related conversations. In an agent environment, ambiguous invocation can route unrelated user requests into this skill, leading to incorrect guidance, prompt-scope confusion, or unintended influence over system behavior and design decisions.

VirusTotal

64/64 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.