ad-creative
Warn
Audited by Gen Agent Trust Hub on Oct 26, 2026
Risk Level: MEDIUMEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTIONDATA_EXFILTRATION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill provides detailed instructions for installing the Voicebox tool, which involves cloning a repository from a third-party GitHub account (jamiepine/voicebox) and executing a local build process using
make. - [REMOTE_CODE_EXECUTION]: The skill documents the use of external generative AI APIs (Gemini, Replicate, Ideogram) and provides shell command examples for interacting with these services via
curl. - [COMMAND_EXECUTION]: The instructions include examples of executing local scripts (
node tools/clis/google-ads.js) and command-line tools for video rendering (npx remotion render) and software setup (make setup). - [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from multiple sources to iterate on ad copy, creating a surface for injection attacks.
- Ingestion points: Performance data provided via CSV files, text pastes, or API outputs, and product marketing context files (
.agents/product-marketing.md). - Boundary markers: None identified; instructions do not specify the use of delimiters or warnings to ignore instructions embedded in the analyzed data.
- Capability inventory: The skill uses tools to read and write to advertising platforms (Google Ads, Meta Ads, LinkedIn Ads, TikTok Ads).
- Sanitization: There is no evidence of sanitization or validation of the input data before it is interpolated into the agent's prompts for creative generation.
- [DYNAMIC_EXECUTION]: The skill utilizes Remotion for video production, which involves executing React and TypeScript code at runtime to render video assets from data templates.
- [DATA_EXFILTRATION]: The skill interacts with several well-known external AI services, including Google Gemini, OpenAI, ElevenLabs, and Ideogram, to generate images, video, and audio assets.
Audit Metadata