Manim Agent

Use this skill to run the full Manim Agent production workflow, not a simplified one-off Manim snippet. Locate the local repository from MANIM_AGENT_HOME, the current workspace, or a user-provided path; if it is missing, clone https://github.com/gqy20/manim-agent.git before running project commands.

Operating Mode

1. Clarify the requested output only when needed: topic, target duration, audience, voice/TTS need, quality level, and final file path.
2. Run scripts/check_manim_agent_env.py before the first real render in a session, or whenever a failure suggests missing dependencies.
3. Prefer the CLI path for direct video delivery. Use the Web path only when the user asks for task history, SSE progress, browser UI, or backend persistence.
4. Preserve the repository pipeline: planning, implementation, render resolution/review, narration, TTS, and mux. Do not replace it with a handmade scene.py unless the user explicitly asks for a raw Manim scene.
5. Produce concrete artifacts: final MP4, generated scene/code location, logs or error summary, and the command used.

Required Interfaces