agentleFS
Edzard9Edzard9/stash/skills/video/manim-explainer/SKILL.md

manim-explainer

Skill0 starsChanged 8 days agoMITView on GitHub

Create short math and science explainer animations with Python Manim: from a concept to a scene plan, a minimal project, a render command, and a verified output file. Use when the user wants a Manim animation, a math explainer video, or says "animasi manim", "buatkan animasi konsep", "manim explainer", "animate this concept". Do NOT use for general web-animation videos (that is html-video) or interactive figures.

  • Installs packages

1 other project uses a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

2 rules. About 0 are common, 1 is distinctive, and 1 is very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈1

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 1

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Concept to beats: Write the explainer as 3-6 beats, one idea per beat, each with the sentence it visualizes.

    line 17
  2. Very specific:

    Keep scenes.py self-contained: no shared utility modules until the second video exists.

    line 54

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Manim Explainer

Manim rewards the same discipline as every render pipeline in this repo: plan the scenes on paper first, render low quality early, and verify the file, not the feeling. A Manim scene that "ran without errors" can still be a black screen with the text off-frame; only the output frames tell the truth.

Workflow

  1. Concept to beats. Write the explainer as 3-6 beats, one idea per beat, each with the sentence it visualizes. If a beat needs two ideas, it is two beats. This list is the storyboard; no code before it exists.
  2. Scene per beat or one scene with timed sections. For short explainers (under 60 seconds), one Scene class with a method per beat keeps timing and cleanup manageable.
  3. Prefer geometry over typeset math when the environment is bare. Text and shapes render anywhere; heavy TeX layouts need a full LaTeX install that small servers often lack. Use plain text objects for labels and numbers unless TeX is confirmed present, and confirm it before writing the scene, not after the failure. The same bareness applies to installation: Manim's cairo dependency builds from source under pip, so on a server with no C compiler pip install manim fails. On such a machine use a conda-forge environment (prebuilt binaries) instead. Check for a compiler before choosing the install route, not after the failed build. (Learned on a real render, 8 November 2026; see examples/.)
  4. Render low first. -ql (480p) for the timing pass, watch it, fix pacing, then -qm or -qh for the delivery render. A pacing mistake found at low quality costs a minute; at high quality it costs the evening.
  5. Verify the artifact. Probe the MP4 (duration, resolution, fps) and extract frames at the first, middle, and last beats. Each frame must show its beat's content, composed inside the frame. A scene whose text overflows is a failed render with a successful exit code.

Minimal project shape

project/
  scenes.py        one file, one Scene subclass per video
  render.sh        the exact commands, low and high quality
  out/             rendered files (the deliverables live here)

Keep scenes.py self-contained: no shared utility modules until the second video exists. Name the output file after the concept, not the scene class.

Pacing rules

  • A beat holds for at least as long as its sentence takes to read aloud, plus a breath. Animations that outrun reading teach nothing.
  • One transformation at a time. Two objects moving while text appears is three things to watch, which is zero things learned.
  • The final frame is a summary state worth screenshotting: the completed diagram, the final value, the whole picture assembled.

Failure notes

  • Missing LaTeX is the most common environment failure; see rule 3.
  • Long renders belong on the bigger machine, supervised like any long job, with the log checked for the render's own progress lines rather than the process merely existing.
  • Fonts: stick to fonts present on the render machine, and verify the extracted frames, because a fallback font changes line breaks and can push text out of frame silently.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.