animation-storyboard
Skill0 starsChanged 2 months agoMITView on GitHub
Turns a brief or message into a validated storyboard.json for a silent-autoplay vector explainer - hook, beats, on-screen text cards, frame timing with reading holds, transitions, CTA, length budget. Use when the user asks to storyboard, script, outline or plan the beats of a short animation or LinkedIn video, or to validate or repair a storyboard.json. Not for screen-recorded walkthroughs.
In this repository10 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- add-skillSkill
- publish-repoSkill
- animation-export-presetsSkill
- animation-frame-qaSkill
- smooth-motion-choreographySkill
- storytelling-animationSkill
- vector-asset-sourcingSkill
- vector-scene-layoutSkill
The file
We split this skill into the rules it gives an agent
3 rules. About 0 are common, 1 is distinctive, and 2 are 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.
2
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 12
No engine, no code: the output is a JSON file that passes the bundled validator.
- Very specific: line 16
Use when the user wants the story, script, beats, scene list, or timing for a short animation (LinkedIn post, slide intro, product story), …
- Very specific: line 20
Do NOT use for: laying out or drawing scenes (vector-scene-layout), animating (smooth-motion-choreography), rendering …
≈ “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.
Files
animation-storyboard
Turns a brief into storyboard.json, the timing contract every later step reads.
It fixes the failure where an agent writes paragraph-length "captions", packs three
ideas into one scene, and produces beats too short to read on a muted phone feed.
No engine, no code: the output is a JSON file that passes the bundled validator.
When to use / when NOT to use
Use when the user wants the story, script, beats, scene list, or timing for a short animation (LinkedIn post, slide intro, product story), or wants an existing storyboard.json fixed or validated.
Do NOT use for: laying out or drawing scenes (vector-scene-layout), animating
(smooth-motion-choreography), rendering (animation-export-presets), reviewing a
render (animation-frame-qa), narrated or voice-over videos (this contract has no
audio track), or storyboarding a screen-recorded product walkthrough.
Workflow
- Read the brief; ask only what is missing. The checklist is in
references/story-rules.md(belief/action after watching, where it is watched, length, verbatim sentences, brand tokens, loop or once, CTA). Default target:linkedin-4x5at 30 fps unless told otherwise. - Write the governing idea in one sentence. Then choose the arc: problem → shift → outcome → CTA for explainers; a single card plus one motion for loops. Pick the beat count from the length table in the reference.
- Draft beats. For each beat write
text(≤ 8 words, a headline, not a subtitle),visual(what the storyboard still shows),motion(intent only), and the handofftransitioninto the next beat. Give every text cardduration_s ≥ 0.7 + overlaps + max(1.33, 0.35 × words, chars / 20)(an 8-word card needs 3.5 s before overlaps). The hook carries text; the CTA beat holds still. Schema:references/storyboard-schema.md; full example:assets/storyboard.example.json. - Validate and fix until it exits 0:
Fix rushed beats by cutting words or moving seconds between beats before lengthening the video; warnings about the recommended length are real trade-offs, say which way you resolved them.python3 "${CLAUDE_SKILL_DIR}/scripts/validate_storyboard.py" storyboard.json - Hand off. Print the validator summary (total frames, per-beat start and hold
frames) and the file path. Downstream steps must use the same per-beat frame
rounding (
timeline.tsin the layout skill does); the summary is the schedule of record. Next:vector-scene-layout.
Output spec
storyboard.json(contract v1) that passesvalidate_storyboard.pywith 0 errors.- The one-sentence governing idea and any brief facts you assumed, stated in the reply.
- Total duration inside the target budget, or an explicit note on why it is longer.
Gotchas
- Autoplay is muted on LinkedIn and 92% of mobile viewers watch without sound: text is the narration. A beat without text must still make sense as a still image.
- Reading speed is the clock (BBC: 0.33–0.375 s per word; Netflix 20 characters/s; DCMP minimum 40 frames). An 8-word card needs 2.8 s of stillness, so a 3.5 s beat with no overlaps; every overlap eats into that.
- Overlapping transitions subtract from the total: two 4 s beats with a 0.5 s fade play for 7.5 s. The validator does this arithmetic; do not hand-sum durations.
- One metaphor per video. If the visuals switch metaphor at the "shift" beat, the storyboard is two videos.
- Loops: the last beat hands off to the first (fade or matching pose, never a cut), and the wrap overlap costs readable time on the hook too.
- Blog posts and decks are not storyboards: reduce to one idea and 4–8 beats, and list what was dropped instead of transcribing.
- Durations are quantised to whole frames per beat (nearest, halves up) and the schedule is summed in frames; 0.1 s steps are exact at 30 fps, 0.25 s steps are not (7.5 frames), so the validator warns and the summary is the truth.
Pointers
scripts/validate_storyboard.py— the contract;--jsonprints the summary for downstream tools,--self-testproves the rules.references/storyboard-schema.md— fields, timing math, minimal example.references/story-rules.md— arcs, platform facts with sources, reading-speed policy, beat counts by length, brief checklist.assets/storyboard.example.json— a validated 546-frame (18.2 s), five-beat example.
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
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.

