ad-creative
Pass
Audited by Gen Agent Trust Hub on Mar 10, 2026
Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface by ingesting ad performance data from external sources such as CSV files, manual pastes, or API outputs to influence creative generation. 1. Ingestion points: Performance data is processed in the Mode 2: Iterate from Performance Data workflow (SKILL.md). 2. Boundary markers: No explicit delimiters or instructions to ignore instructions within the performance data are provided. 3. Capability inventory: The skill utilizes command execution (node tools, ffmpeg) and network requests to external AI APIs (SKILL.md, references/generative-tools.md). 4. Sanitization: There are no instructions for sanitizing or validating the input data before processing.
- [COMMAND_EXECUTION]: The documentation includes examples of executing local Node.js scripts for pulling ad performance data and using system utilities like ffmpeg and make for creative production and tool setup.
- [EXTERNAL_DOWNLOADS]: The skill references external resources including the Remotion framework via npx and the Voicebox repository via git clone. These are well-known services and tools used for creative production.
Audit Metadata