shihabshahrier/manim-coding-skill / skills/test-manim
test-manim
VerifiedCreate teaching-quality STEM animations using ManimGL. Generates narrative-driven animations as 15-second chunks merged into a final video. Configurable topic, duration, and model name. Works with Claude Code and opencode across any model. Invoke: /test-manim "topic" [--duration 30|60|90|120] [--model name] Default duration: 30s. Primary goal: great storytelling + visual clarity.
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 9.8 KB |
| references/manim-gl-api.md | markdown | 6.0 KB |
| references/storytelling.md | markdown | 4.7 KB |
references/storytelling.md
View on GitHub# STEM Animation Storytelling Guide
Core principle: the viewer should feel understanding arrive, not just receive information.
---
## The 3-Act Structure (mandatory)
### Act 1 — The Question (first chunk)
Don't start with a definition. Start with a phenomenon.
- Show the effect before naming the cause
- Pose a visual question the viewer wants answered
- Good: ball in motion, pendulum swinging, wave propagating
- Bad: "Today we will learn about Newton's First Law"
Opening frames: concrete, specific, surprising.
### Act 2 — The Build (middle chunks)
One concept per chunk. Strict.
- Introduce one new idea per chunk
- Each chunk should make the PREVIOUS chunk more meaningful
- Use "yes, but why?" structure: show result → reveal mechanism
- Never skip the intuitive step to get to the formal step
Build order: phenomenon → intuition → formalism → consequence
### Act 3 — The Synthesis (last chunk)
Return to the opening image, transformed by understanding.
- Re-show the Act 1 visual, now annotated with what was learned
- "Before we knew why. Now we do."
- End on a held frame: the key insight visible on screen
- Last 3–5 seconds: static, clean, memorable
---
## Pacing Rules
| Moment type | Pacing |
|-------------|--------|
| New object appears | slow — 1.5–2.0s |
| Familiar object moves | fast — 0.5–1.0s |
| Equation appears | slow — 1.5–2.5s |
| Key insight holds | still — wait(3) |
| Transition between ideas | fade — 0.5s |
| Closing frame | still — wait(4) |
Don't rush insights. Viewers need time to connect dots.
Slow = deliberate, not boring.
---
## Visual Language Rules
### Color is meaning, not decoration
Pick 3–4 colors max. Each color = one concept. Never reassign.
Example for Newton's Laws:
- BLUE = object / mass
- YELLOW = force vector
- RED = acceleration
- WHITE = labels, equations
If chunk 3 uses BLUE for velocity instead of mass → confusion.
### Motion direction is semantic
- Right → positive, forward, future, increase
- Left → negative, backward, past, decrease
- Up → increase, gain, higher
- Down → decrease, loss, gravity
Don't violate these unless you're making a point about violation.
### Labels arrive after the visual
Wrong order: show equation → animate what it means
Right order: animate the concept → reveal the equation that describes it
Viewer thinks: "Oh, that's what F=ma means" not "let me remember what F=ma says"
### Camera movement = perspective shift
Move camera only to:
1. Zoom in: "look closely at this detail"
2. Zoom out: "see the bigger picture"
3. Pan: "here's something related, spatially"
Never zoom just for visual variety. Each camera move should feel like turning a page.
---
## Chunk Design Checklist
Before writing code for a chunk:
- [ ] What is the ONE concept this chunk teaches?
- [ ] What is the FIRST thing the viewer sees? (opening frame)
- [ ] What is the SURPRISE or INSIGHT moment? (at ~10s)
- [ ] What does the viewer understand at second 15 that they didn't at second 0?
- [ ] Does this chunk reference the color contract?
- [ ] Does this chunk end cleanly (FadeOut or held frame)?
---
## Common Anti-Patterns
**Anti: Definition dump**
Opening with `F = ma` text before showing anything.
Fix: Show a ball accelerating under force first.
**Anti: Too many elements**
5 objects appear simultaneously.
Fix: Introduce one, establish its meaning, then introduce the next.
**Anti: Fast equations, slow everything else**
Equation flashes onscreen in 0.3s.
Fix: Write equations with `run_time=2.0` minimum.
**Anti: Color chaos**
Using BLUE for two different concepts in the same animation.
Fix: Assign colors in SCENE_PLAN.md and never deviate.
**Anti: No resolution**
Last chunk introduces a new concept instead of synthesizing.
Fix: Last chunk must reference chunk 1 visuals.
**Anti: Wall of text**
Long sentences as `TexText` objects.
Fix: Max 8 words per text object. Use multiple lines if needed.
---
## Strong Opening Patterns
These chunk-1 openings consistently work:
1. **Object in motion**: Show before force is introduced. "Why does this keep moving?"
2. **Pattern without explanation**: Show a Fourier approximation building. "How does this work?"
3. **Two contrasting cases**: Ball rolls forever on ice vs. stops on carpet. "What's different?"
4. **Scale reveal**: Zoom from macro to micro (or reverse). "What's happening here?"
5. **Paradox setup**: Show something that seems to violate intuition. "Wait, why?"
---
## Audio Note
Skill produces silent video. When audio is added later:
- Leave 0.5s silence at start of each chunk (room for voice-over sync)
- Hold frames at insight moments 3–5s minimum (narration landing time)
- Keep text on screen 1.5× longer than time to read it
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| finishing-a-development-branchobra/superpowersUse when implementation is complete, all tests pass, and you need to decide how to integrate the work | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| systematic-debuggingobra/superpowersUse when encountering any bug, test failure, or unexpected behavior, before proposing fixes | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| test-driven-developmentobra/superpowersUse when implementing any feature or bugfix, before writing implementation code | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| skill-creatoranthropics/skillsCreate new skills, modify and improve existing skills, and measure skill performance. Use when users want to create a skill from scratch, edit, or optimize an existing skill, run evals to test a skill, benchmark skill performance with variance analysis, or optimize a skill's description for better triggering accuracy. | Stars: 198k | Updated | Claude CodeCodex | Verified | Install |