The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs the agent to download, install, and execute software from untrusted third-party sources. In references/imessage-video-ads.md, it recommends installing an external pipeline via npx gooseworks install --all. In references/generative-tools.md, it provides instructions to clone and build an open-source voice tool from a personal GitHub repository: git clone https://github.com/jamiepine/voicebox.git. In references/motion-video-ads.md, it references a third-party repository Bomx/super-video-maker-skill for creative recipes.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted external data which could contain malicious instructions meant to override agent behavior. Ingestion points: The 'Grounded Inputs' workflow in SKILL.md requires reading customer reviews from Trustpilot, Amazon, and G2, as well as ad comments. Boundary markers: The instructions do not define delimiters or specific 'ignore' directives to prevent the agent from following instructions embedded in the review or comment data. Capability inventory: Across SKILL.md and referenced files, the agent has access to ad platform tools (e.g., google-ads campaigns create) with write permissions, and can write output files locally. Sanitization: Validation focuses on platform character limits rather than instruction filtering or content sanitization.
[COMMAND_EXECUTION]: The skill provides numerous shell command patterns for the agent to follow. SKILL.md includes examples for running local Node.js CLI tools (node tools/clis/google-ads.js). references/imessage-video-ads.md and references/generative-tools.md provide commands for package installation and repository setup.

ad-creative