The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill's core functionality relies on ingesting external data from customer reviews and ad comments (found in inputs/reviews/ and inputs/comments/) to ground its creative generation. This represents a vulnerability surface where malicious instructions embedded in customer-provided text could influence the agent's behavior or final ad output.
Ingestion points: Customer reviews and community comments stored in the inputs/ directory (SKILL.md).
Boundary markers: Absent; there are no instructions to use specific delimiters or protective framing when processing customer text.
Capability inventory: The skill possesses the ability to write to the file system (batch outputs), execute commands via ad platform CLIs (google-ads, meta-ads), and make outbound network requests to various generative AI APIs.
Sanitization: Absent; the instructions do not specify sanitization or validation of customer-supplied text before it is included in prompts or ad drafts.
[DYNAMIC_EXECUTION]: The skill facilitates the creation and execution of dynamic content to support ad previews and automated video production. This includes a "Creative Review Page" that uses an HTML/JS template (assets/creative-review-template.html) to render ads from a JSON data object. Furthermore, the skill describes a production pipeline involving Playwright for headless browser recording, ffmpeg for media assembly, and Remotion for rendering React-based video compositions from templates.
[EXTERNAL_DOWNLOADS]: To perform its primary duties, the skill provides instructions for and expects integration with a wide array of external AI services and tools. These include well-known platforms like Google Gemini, OpenAI TTS, and ElevenLabs, as well as specialized services and repositories such as Flux (via Replicate), Ideogram, Kling, Runway, Seedance, and third-party utilities like the gooseworks ad tool and the voicebox voice synthesis studio.

ad-creative