ad-creative
Pass
Audited by Gen Agent Trust Hub on Jul 27, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: No malicious instructions, obfuscation, or sensitive data access patterns were found in the skill or its associated reference files. The skill focuses on legitimate performance marketing workflows.
- [PROMPT_INJECTION]: The skill contains a vulnerability surface for indirect prompt injection, as it is designed to ingest and analyze external performance data (CSV, paste, or API output). This is a functional requirement for its intended purpose of creative iteration.
- Ingestion points: User-provided ad performance metrics (SKILL.md).
- Boundary markers: None explicitly mentioned in the skill for handling untrusted data.
- Capability inventory: Access to platform-specific tools (
google-ads,meta-ads,linkedin-ads,tiktok-ads) for data retrieval and campaign management. - Sanitization: Not explicitly implemented in the skill instructions; relies on the agent's internal guardrails.
- [REMOTE_CODE_EXECUTION]: The skill references the Remotion framework for programmatic video generation and mentions using AI to generate React code for templates. These are standard developer operations for the defined use case and do not constitute an execution of untrusted remote scripts.
- [EXTERNAL_DOWNLOADS]: The skill references several well-known AI services (Google Gemini, ElevenLabs, OpenAI, Replicate, ByteDance) and provides instructions for installing the Voicebox open-source tool from GitHub. These are identified as trusted or well-known services and do not pose a security risk.
Audit Metadata