ClawHub
Back to skill

Security audit

多代理开发框架

Security checks across malware telemetry and agentic risk

Overview

The skill is a coherent multi-agent coding workflow, but its broad activation and inconsistent disclosure around write/exec authority warrant Review before installation.

Install only if you want an agent to actively execute coding plans in a repository, including reading files, editing files, running commands/tests, coordinating subagents, and possibly creating commits. Review the pure-Markdown classification mismatch and avoid using callback URLs or API keys unless you understand exactly what data the agent will send.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (4)

Intent-Code Divergence

Medium
Confidence
95% confidence
Finding
The skill simultaneously advertises itself as a pure Markdown skill while declaring executable capabilities earlier in the manifest (`exec`, `write`, etc.). This inconsistency can mislead users and host platforms about the real execution surface, reducing informed consent and increasing the chance that code execution or file modification occurs under a lower-risk label.

Vague Triggers

Medium
Confidence
91% confidence
Finding
The trigger language is broad enough to match ordinary development requests, which can cause the skill to activate in contexts where the user did not intentionally request this higher-privilege workflow. Because the skill includes command execution, file writes, and orchestration behavior, accidental invocation increases the risk of unintended actions.

Vague Triggers

Medium
Confidence
93% confidence
Finding
The example phrase is generic and likely to collide with normal conversation about software development. In a skill that can read files, write files, and execute commands, such collision can lead to unintentional privilege use and execution of complex workflows without clear user intent.

Missing User Warnings

Medium
Confidence
96% confidence
Finding
The skill describes file writes, command execution, and callback URL handling without presenting a prominent upfront warning about these powerful behaviors. Users may engage the skill expecting planning assistance, while the skill can perform state-changing local actions and potentially send data to external endpoints, which materially raises security and privacy risk.

VirusTotal

VirusTotal findings are pending for this skill version.

View on VirusTotal

Static analysis

No suspicious patterns detected.