ClawHub
Back to skill

Security audit

Atlas Cro Lite

Security checks across malware telemetry and agentic risk

Overview

This is a prompt-only landing-page conversion audit skill with no executable code or privileged access, though it has broad wording and one unrelated paid DeFi upsell line.

Install this only if you want a lightweight CRO/copywriting prompt helper. Review generated reports before sharing because they may include Atlas promotional text, and be cautious using it on sensitive, regulated, authenticated, or proprietary pages. Treat the DeFi upsell and version mismatch as publisher cleanup issues, not evidence of malicious behavior.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • 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
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (3)

Intent-Code Divergence

Medium
Confidence
96% confidence
Finding
The upgrade section advertises 'advanced DeFi modules' inside a CRO/landing-page optimization skill, which is a clear capability mismatch. This can mislead users about the scope of the package, normalize unrelated high-risk financial/security workflows in a marketing context, and indicates possible template contamination or deceptive upsell content.

Vague Triggers

Medium
Confidence
82% confidence
Finding
Claiming the skill works on 'any' landing page and for generic CRO/copywriting tasks creates overly broad activation boundaries. In agent environments, this can cause the skill to be invoked outside its intended context, leading to inappropriate handling of pages with regulated, sensitive, or out-of-scope content.

Vague Triggers

Medium
Confidence
84% confidence
Finding
The instruction to 'quickly audit any landing page' lacks guardrails on when the skill should or should not be used. Without boundaries, downstream agents may apply it indiscriminately to sensitive, proprietary, or unsuitable targets, increasing the risk of scope creep and unsafe automation.

VirusTotal

64/64 vendors flagged this skill as clean.

View on VirusTotal