facebook-ads-campaign
Pass
Audited by Gen Agent Trust Hub on Sep 17, 2026
Risk Level: SAFE
Full Analysis
- [PROMPT_INJECTION]: The skill includes robust defenses against prompt injection. It explicitly instructs the agent to treat all external data (briefs, websites, exports) as untrusted data and never as instructions. It specifically warns against following URLs found within fetched content and mandates that any instruction found inside content must be reported as a finding rather than obeyed.
- [INDIRECT_PROMPT_INJECTION]: The skill acknowledges the risk of processing untrusted data and implements strict boundaries. It requires explicit human confirmation ('publish') before any action that involves spending money, preventing automated exploitation of the agent's write access to ad accounts.
- [DATA_EXFILTRATION]: The skill includes comprehensive guidelines for handling sensitive data. It instructs the agent to redact credentials found in inputs, avoid persisting raw personal data, and minimize the data written to state files. It also forbids the use of protected categories (race, religion, etc.) for targeting or segmentation.
- [COMMAND_EXECUTION]: No unauthorized or dangerous command execution patterns were detected. The skill uses standard MCP (Model Context Protocol) for interacting with ad accounts, which is a controlled environment.
- [DYNAMIC_EXECUTION]: The skill does not use dynamic code generation or unsafe deserialization. It includes instructions for HTML-escaping any content interpolated into emitted markup to prevent XSS.
Audit Metadata