The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill recommends the installation and use of third-party tools from unverified sources. Specifically, references/imessage-video-ads.md suggests using npx gooseworks install --all, and references/generative-tools.md instructs the user to clone the voicebox repository from github.com/jamiepine/voicebox.git. These sources are not recognized as trusted organizations or well-known services.
[REMOTE_CODE_EXECUTION]: The inclusion of instructions to install the gooseworks package and execute code-based pipelines (e.g., the Playwright + ffmpeg pipeline described in references/imessage-video-ads.md) involves downloading and running code from external, unverified repositories.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data which creates a vulnerability surface.
Ingestion points: The agent is instructed to read customer reviews and ad comments from inputs/reviews/ and inputs/comments/ (SKILL.md).
Boundary markers: No specific boundary markers or instructions to ignore embedded commands within the input data are defined.
Capability inventory: The skill uses network-enabled CLI tools for advertising platforms (google-ads, meta-ads, linkedin-ads, tiktok-ads), performs file system writes in the outputs/ directory, and executes shell commands including curl, node, and ffmpeg (SKILL.md, references/generative-tools.md).
Sanitization: While references/creative-review-page.md suggests escaping < as \u003c within JSON objects to prevent script breakout in HTML templates, there are no instructions to sanitize or filter the raw input data from reviews or comments before they are interpolated into prompts.
[DYNAMIC_EXECUTION]: The skill utilizes dynamic content generation frameworks like Remotion (React-based video rendering) and automated browser-based recording via Playwright (references/generative-tools.md, references/imessage-video-ads.md), which involve executing scripts and rendering logic at runtime to produce creative assets.

ad-creative