The Agent Skills Directory

[COMMAND_EXECUTION]: The script lesson-screen.mjs executes shell commands to start a local server and launch web browsers (Chrome, Chromium, or Edge) for rendering and screenshots. This functionality is scoped to the primary purpose of providing a visual lesson interface.
[EXTERNAL_DOWNLOADS]: The skill fetches lesson data, practice questions, and assets from the author's public GitHub repository (rohitg00/ai-engineering-from-scratch). This is the intended behavior for retrieving course curriculum.
[INDIRECT_PROMPT_INJECTION]: The tool processes local lesson files such as docs/en.md and quiz.json to generate interactive HTML screens.
Ingestion points: Files are read from the local repository or the .lesson-screen session directory.
Boundary markers: Standard HTML templates are used; no specific 'ignore instruction' markers are present for the LLM itself.
Capability inventory: The skill can spawn subprocesses (browsers) and write files to the session directory.
Sanitization: Content is escaped using an internal escapeHtml function and markdown parsing logic before being embedded in templates. A restrictive Content Security Policy (CSP) is applied to the served pages to prevent unauthorized network requests.

lesson-screen