Ad Manager

• Outcome: Review
• Latest audit: May 29, 2026 at 9:25 PM UTC
• Version:
  1.2.4

Overview

This ad-management skill matches its advertised purpose, but it can launch or change paid campaigns from broad natural-language requests without a clear final confirmation step.

Install only if you are comfortable giving an agent advertising-management authority. Use the test agent first, connect real ad accounts only deliberately, and require a final human approval before creating campaigns, uploading creatives, changing status, or reallocating budget.

SkillSpector

By NVIDIA

Vulnerability Patterns

• Trigger Abuse: Overly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
• Prompt Injection: Instruction Override, Hidden Instructions, Exfiltration Commands
• Data Exfiltration: External Transmission, Env Variable Harvesting, File System Enumeration
• Privilege Escalation: Excessive Permissions, Sudo/Root Execution, Credential Access
• Supply Chain: Unpinned Dependencies, External Script Fetching, Obfuscated Code

Findings (2)

Vague Triggers

Medium

• Confidence: 88% confidence
• Finding: The example trigger phrase for creating ads is very broad and presented as a natural-language command without any invocation boundary, namespace, or confirmation step. This can cause accidental activation during ordinary conversation or when discussing ad planning, leading the agent to initiate campaign-management behavior unexpectedly in a business-sensitive context.

Vague Triggers

Medium

• Confidence: 85% confidence
• Finding: The analysis example is similarly ambiguous and can overlap with ordinary requests like asking for a summary of recent ad performance. In a skill that handles campaign optimization and reporting, ambiguous activation may expose sensitive marketing data or trigger automated analysis workflows when the user only intended general discussion.

VirusTotal

66/66 vendors flagged this skill as clean.