The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from user-supplied app screenshots and metadata retrieved from the App Store via scripts/fetch_listing.py. This content is used to influence the generated design styles and marketing copy.
Ingestion points: User screenshots provided during intake and app metadata fetched from the Apple App Store API.
Boundary markers: No explicit delimiters are used to separate untrusted external content from agent instructions during storyboard generation.
Capability inventory: The skill has the ability to write to the local filesystem, perform network requests to external AI APIs, and execute bundled Python and Node.js scripts.
Sanitization: No explicit sanitization or filtering of the processed metadata or screenshot analysis results was identified.
[EXTERNAL_DOWNLOADS]: The skill retrieves configuration and assets from well-known services and trusted repositories.
Fetches device frames from the official fastlane/frameit-frames repository on GitHub.
References the Three.js library and Simple Icons set through the jsdelivr.net CDN for rendering tasks.
Communicates with official endpoints for OpenAI, Google Gemini, and the Apple App Store for image generation and metadata retrieval.
[COMMAND_EXECUTION]: The skill utilizes several internal utility scripts to perform its core functions, which is standard for high-complexity AI agent tasks.
scripts/fetch_listing.py: Downloads screenshots and metadata from official store listings.
scripts/gen_image.py: A wrapper for interacting with OpenAI and Google image generation APIs using environment-stored credentials.
scripts/export.js: Uses the Playwright framework to render HTML screenshot panels and export them as finalized PNG/JPEG files.
scripts/chroma_key.py: A Python script using standard libraries for background removal on generated assets.

app-store-screenshots