ClawHub
Back to skill

Security audit

迭代优化器

Security checks across malware telemetry and agentic risk

Overview

This is an instruction-only code improvement skill with no install actions, credentials, persistence, or hidden execution.

Before installing, understand that ordinary requests to improve or optimize code may activate this skill. It does not run code or install anything by itself, but you should still review any proposed edits or commands and avoid sharing source files that contain secrets.

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
90% confidence
Finding
The activation phrases are very broad (e.g. general requests to improve or optimize code) and can match ordinary developer prompts unintentionally. This increases the chance the skill auto-triggers in contexts the user did not explicitly intend, causing unsolicited workflow changes or repeated optimization behavior.

VirusTotal

65/65 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.