The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill generates video scripts and assets by analyzing content captured from external applications (iOS, Android, Web, and macOS).
Ingestion points: Content is ingested from user-provided briefs and automated screenshots/recordings of external applications as defined in references/capture.md.
Boundary markers: No explicit delimiters or instructions are provided to the agent to ignore potentially malicious commands embedded within the UI text of the applications being captured.
Capability inventory: The skill has the capability to write files (writeFile), execute shell commands (spawn for ffmpeg, adb, xcrun), and control browser environments via Playwright.
Sanitization: There is no evidence of sanitization or filtering of text extracted from application UIs before it is interpolated into the generated video composition.
[EXTERNAL_DOWNLOADS]: The skill automates the installation of the playwright package and its associated Chromium browser engine using npm install and npx playwright install. These components are fetched from trusted Microsoft-maintained registries to provide the rendering environment.
[COMMAND_EXECUTION]: The skill utilizes several command-line tools for media processing and device interaction, including ffmpeg for video encoding, xcrun simctl for iOS simulation, and adb for Android interaction. These tools are invoked via the spawn method in engine/render.mjs and through explicit instructions in references/capture.md.
[DYNAMIC_EXECUTION]: The core rendering logic involves generating a motion.html composition which is then executed within a headless browser context. The engine/render.mjs script uses Playwright's page.evaluate to programmatically drive the animation sequence, representing a controlled dynamic execution pattern necessary for the skill's functionality.

motion-graphics