Manim Video Skill

manim-video is a bundled CoWork OS skill for planning, scaffolding, and optionally rendering technical explainer videos with Manim Community Edition.

It is designed for:

• animated math walkthroughs
• equation derivations
• algorithm visualizations
• technical concept explainers
• architecture build-up animations
• animated data stories
• 3Blue1Brown-style educational videos

It is not the right tool for:

• live-action editing
• stock-footage montage
• generic marketing motion graphics
• static diagrams or slide decks
• brainstorming-only requests with no renderable output

What The Skill Does

Compared with a plain imported SKILL.md, CoWork’s bundled manim-video integration adds a stronger local workflow:

• dependency preflight through resources/skills/manim-video/scripts/setup.sh
• deterministic project scaffolding through resources/skills/manim-video/scripts/bootstrap_project.py
• workspace-local outputs instead of an ephemeral prompt-only answer
• explicit draft-vs-production render flow
• run artifacts for project review and handoff

The skill scaffolds a Manim project with:

• plan.md
• script.py
• concat.txt
• render.sh
• optional voiceover.md

It also expects run artifacts under the current task artifact directory:

• project-manifest.md
• render-checklist.md
• review-notes.md

How To Use It

manim-video is a built-in bundled skill. There is nothing to install from the Skill Store.

The easiest way to use it is to ask directly in natural language, for example:

Create a 3Blue1Brown-style Manim video explaining gradient descent.

Build a Manim animation that visualizes Dijkstra's algorithm step by step.

Use the manim-video skill to create an animated equation derivation for the chain rule.

Good requests usually include:

• the topic
• the audience
• the target length
• whether voiceover should be included
• whether you want only scaffolding or actual rendering

Example:

Use the manim-video skill to create a 75-second concept explainer for gradient descent aimed at software engineers. Scaffold the full project in this workspace and render only draft quality first.

Invocation Model

manim-video follows CoWork’s additive skill runtime.

• The original task stays canonical.
• The skill adds execution context and scoped directives.
• It does not replace the user’s task with a synthetic prompt.

See Skills Runtime Model.

Parameters

The bundled manifest supports these inputs:

• topic: what the animation explains or visualizes
• mode: auto, concept-explainer, equation-derivation, algorithm-visualization, data-story, architecture-diagram, paper-explainer, or 3d-visualization
• audience: target audience for pacing and explanation depth
• target_length_seconds: approximate runtime
• output_dir: workspace-relative or absolute output directory
• voiceover: auto, on, or off

If output_dir is omitted, the skill defaults to a local manim-video-project directory in the current workspace.

Project Workflow

The skill’s expected flow is:

1. Run the setup preflight.
2. Read the bundled guidance and troubleshooting references.
3. Bootstrap the project skeleton.
4. Write or update plan.md and script.py.
5. Render draft quality first if execution is requested and dependencies are satisfied.
6. Review clarity, pacing, and scene transitions.
7. Only then move to production quality.

The generated script.py uses one Scene subclass per beat and shared constants at the top of the file.

Dependencies

The skill checks for:

• Python 3.10+
• Manim Community Edition in the active Python environment
• ffmpeg
• a LaTeX engine such as pdflatex
• dvisvgm when available

You can run the same preflight manually:

bash resources/skills/manim-video/scripts/setup.sh

On a machine where Manim is missing, the skill still remains useful for planning and project scaffolding, but draft rendering will not be available until the dependency is installed.

Generated Files

plan.md

Contains:

• topic and audience
• target runtime
• visual language
• scene breakdown
• pacing notes
• review checklist

script.py

Contains:

• shared palette and typography constants
• one Manim scene class per beat
• scaffolded timing and cleanup patterns
• renderable scene names such as Scene01Hook

render.sh

Provides the standard entry points:

• bash render.sh draft
• bash render.sh production
• bash render.sh still Scene02Intuition

Recommended Prompt Patterns

Use this skill when the output should teach through motion.

Strong fits:

• “Explain backpropagation with animated geometry.”
• “Animate how a binary heap changes during insert and extract-min.”
• “Turn this research paper’s core method into a short visual explainer.”
• “Build an animated architecture diagram showing request flow through our system.”

Weak fits:

• “Edit this podcast clip.”
• “Make a generic product promo video.”
• “Create a static diagram.”

Related Features And Skills

• Features: product-wide runtime and skills overview
• Use Cases: copy-paste prompts that include manim-video
• Use Case Showcase: example workflows powered by the skill
• Skill Store & External Skills: explains why this one is bundled and available immediately
• video-frames: for extracting stills or clips from an existing video, not generating a new animation
• built-in video generation providers: better for model-generated video clips, not deterministic technical animation

Where The Source Lives

Bundled skill files:

• resources/skills/manim-video.json
• resources/skills/manim-video/SKILL.md
• resources/skills/manim-video/references/full-guidance.md
• resources/skills/manim-video/references/troubleshooting.md
• resources/skills/manim-video/scripts/setup.sh
• resources/skills/manim-video/scripts/bootstrap_project.py

Development Notes

When editing the bundled skill itself, run:

python3 -m py_compile resources/skills/manim-video/scripts/bootstrap_project.py
bash resources/skills/manim-video/scripts/setup.sh
npm run skills:check