The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill dynamically generates several TypeScript files including scripts/recording-setup.ts, scripts/playwright-capture.ts, and scripts/interactive-record.ts. It subsequently executes these scripts using npx tsx to manage the recording and setup phases.
[CREDENTIALS_UNSAFE]: The skill captures browser session state (cookies and localStorage) and saves it to a local file at public/recordings/<flow-slug>/storage-state.json. This is used to persist authentication for automated recordings but results in sensitive credential data being stored in the project directory.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from external websites (element text, aria-labels, and navigation URLs) to generate captions and highlights for the tutorial video.
Ingestion points: Browser interaction data captured in scripts/interactive-record.ts and scripts/playwright-capture.ts.
Boundary markers: None identified in the prompt templates.
Capability inventory: File system writing, command execution (npx, ffmpeg, say), and network access via Playwright.
Sanitization: Minimal sanitization; text is trimmed or sliced for length but not validated for prompt injection patterns.
[COMMAND_EXECUTION]: The skill invokes multiple shell commands including npx playwright, npx tsx, ffmpeg, ffprobe, and the macOS say utility to perform automation, video encoding, and text-to-speech generation.

playwright-tutorial-video