Back to skill

Security audit

ai-video-creator-assistant

Security checks across malware telemetry and agentic risk

Overview

This paid video-assistant skill is mostly coherent, but it under-discloses that the user’s video prompt is saved locally with order/payment metadata.

Install only if you are comfortable with a paid workflow that writes local order records, including your video topic or prompt, under ~/.openclaw/skills/orders. Avoid putting confidential, personal, or proprietary details in the initial video description unless the publisher updates the disclosure and retention behavior.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Rogue AgentSelf-Modification, Session Persistence
  • 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
Findings (5)

Intent-Code Divergence

Medium
Confidence
95% confidence
Finding
The script prints a notice claiming that no video content, API keys, or personal data are stored, but the order file written by create_order_file includes the user-provided question field. A video topic or idea description can contain personal, confidential, or commercially sensitive information, so this is misleading data handling and creates unnecessary local persistence risk.

Vague Triggers

Medium
Confidence
74% confidence
Finding
The trigger phrase around '说"帮我做个XX视频"' is broad enough to match normal conversation, which can cause the skill to activate unintentionally. In a paid workflow with credential and filesystem permissions, accidental activation increases the risk of confusing users, initiating payment-related steps, or prompting for sensitive API-key setup when the user did not intend to use this skill.

Natural-Language Policy Violations

Medium
Confidence
83% confidence
Finding
Mandating Chinese-language interaction without user opt-in can mislead or exclude users and may cause them to misunderstand payment, credential, or privacy-related instructions. In this skill, misunderstanding is especially relevant because the workflow involves external API keys and a paid order flow, so language constraints can degrade informed consent.

Missing User Warnings

Medium
Confidence
97% confidence
Finding
The script stores untrusted user-supplied question content into a local JSON order file without a clear and accurate warning to the user. In this skill context, the question is likely to be the creative brief for a video and may include personal data, proprietary marketing plans, or sensitive prompts, so silent persistence increases privacy and data exposure risk.

Session Persistence

Medium
Category
Rogue Agent
Content
return order_data

if __name__ == "__main__":
    parser = argparse.ArgumentParser(description="Create ai-video-creator order")
    parser.add_argument("question", help="Video topic / idea description")
    args = parser.parse_args(); indicator = compute_indicator(SLUG)
    print("=" * 60)
Confidence
82% confidence
Finding
Create ai-video-creator order") parser.add_argument("question", help="Video topic / idea description") args = parser.parse_args(); indicator = compute_indicator(SLUG) print("=" * 60) p

VirusTotal

65/65 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.