claude-video-skills
CLAUDE.md4 starsChanged 4 months agoMITView on GitHub
Video-editing automation for Claude Code. All video work starts at /make-video (.claude/skills/make-video/SKILL.md) — it asks what you're making, scaffolds projects/<name>/, and routes to the right pipeline. Don't pick a sub-skill directly unless the user names it. Per-video runbook: EXECUTION.md. Install + brand setup: SETUP.md. Session state and standing decisions: WORKLOG.md (read first, append after every session). Route C's recurring-series pattern (STYLE.md + shared series_pipeline.sh) and its selectable visual style are documented in EXECUTION.md. Routes A/B (and route C when visual-os…
- Reads credentials
In this repository5 other files it gives its agents.
- make-videoSkill
- motion-design-editSkill
- reel-scriptSkill
- video-edit-longformSkill
- video-editSkill
The file
We split this CLAUDE.md into the rules it gives an agent
1 rule. About 0 are common, 1 is distinctive, and 0 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.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 54
Never auto-run a final render — preview first, explicit user sign-off.
≈ “Common” is an estimate: at least 1 in 5 of the 28,029 CLAUDE.md 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
video-pipeline
Video-editing automation for Claude Code. All video work starts at
/make-video (.claude/skills/make-video/SKILL.md) — it asks what you're
making, scaffolds projects/<name>/, and routes to the right pipeline. Don't
pick a sub-skill directly unless the user names it.
| Making | Route | Engine |
| Talking-head short (9:16) | A | .claude/skills/video-edit — hf_* graphics |
| Long-form (16:9) | B | .claude/skills/video-edit-longform — viz graphics |
| Faceless motion reel (VO only) | C | .claude/skills/motion-design-edit — pr_* graphics, Paper Brand or Visual OS style (per project) |
| Reel script only | D | .claude/skills/reel-script |
| Faceless motion long-form (16:9) | planned, not yet routed | — |
Per-video runbook: EXECUTION.md. Install + brand setup: SETUP.md.
Session state and standing decisions: WORKLOG.md (read first, append
after every session). Route C's recurring-series pattern (STYLE.md +
shared series_pipeline.sh) and its selectable visual style are documented
in EXECUTION.md.
Brand — not one-size-fits-all
Routes A/B (and route C when visual-os is chosen) render in whatever brand
BRAND.md records. BRAND.md does not ship — the "Brand setup" flow
in SETUP.md writes it. Run that flow once, before your first real render, to
replace the shipped example preset with your own colors and fonts — it's a
short conversation (primary color, a second accent, fonts, anything to
avoid), not a config file to hand-fill. Until then the example preset below
is what renders.
- Ships with an example "Visual OS" preset until you customize it: deep navy ink, electric-blue accent, warm punch color, light background; Georgia titles · Inter body · JetBrains Mono peripheral chrome only (eyebrows, tags, footers, annotations).
- Discipline that survives any rebrand: one clear accent color per frame, a punch/execution color distinct from it, peripheral-chrome fonts never bleed into headline/body roles.
- Token files (mirror
BRAND.md's values; edit them to re-skin, never hardcode a one-off color inline):.claude/skills/video-edit/remotion/src/templates/hfBrand.ts·.claude/skills/video-edit-longform/remotion/src/motion/tokens.ts·.claude/skills/motion-design-edit/remotion/src/templates/paperBrand.ts - Route C (
pr_*) is the one exception: it supports two selectable styles — Paper Brand (cream/ink/orange, the original reference-reel look, default) or the Visual-OS-style preset — chosen per project via.pr_style. Paper Brand intentionally does NOT follow your configured brand; it's its own independent look.
Hard rules
- Read
MOTION-STYLE.md(repo root) before authoring any beat plan or animation code — visual concepts not captions, motion density, asymmetry. - Never auto-run a final render — preview first, explicit user sign-off.
- Pass an absolute
OUT_DIRto any render.sh. - Transcribe via
.claude/skills/make-video/scripts/transcribe_assemblyai_bridge.py(needs ONE ofASSEMBLYAI_API_KEY/OPENAI_API_KEYin.env); verify the transcript and hand-patchwords.json. - Finals →
projects/<name>/outputs/, previews →outputs/previews/, intermediates →~/.cache/video-edit/<hash>/. Never write intoraw/. - This machine's ffmpeg may be a minimal build with no libass — captions/ graphics render inside Remotion, never via ffmpeg subtitle burn.
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.

