The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: In the references/generative-tools.md file, the skill provides instructions for the user or agent to clone a repository (https://github.com/jamiepine/voicebox.git) and run build commands (make setup && make dev). This involves downloading and executing arbitrary code from an untrusted source, presenting a supply chain risk.
[COMMAND_EXECUTION]: The skill suggests using several shell-based workflows. This includes using npx to download and execute NPM packages (create-video, remotion) and executing local Node.js scripts (node tools/clis/google-ads.js) to interact with ad platforms.
[EXTERNAL_DOWNLOADS]: The skill facilitates the download of third-party tools and assets, specifically through git clone for the Voicebox repository and npx for Remotion utilities.
[DATA_EXFILTRATION]: The skill includes example curl commands that send content (such as ad scripts and descriptions) to various third-party AI service providers (Google Gemini, ElevenLabs, Cartesia, Ideogram). While these are recognized services, this behavior constitutes sending potentially sensitive product data to external endpoints.
[PROMPT_INJECTION]: The skill is designed to ingest external data, such as marketing context and performance metrics (CSV/API output), and interpolate it into prompts. There is a risk of indirect prompt injection as the skill does not define strict boundary markers or sanitization logic for these external inputs.

ad-creative