Creative Studio Skills
Vasanth19/skills
Skills para produção de vídeo, áudio, IA e automação social
Ver no GitHubBiblioteca de skills Claude Code para produção de vídeo, automação social e mídia com IA. Inclui ferramentas compostas (ffmpeg, áudio, script), integrações com plataformas AI (HeyGen, Gemini, Higgsfield) e pipelines completos para reels, vídeos long-form e gráficos. Gerenciamento otimizado via Paperclip para evitar sobrecarga de tokens.
Instalação
npx convez-ai Vasanth19/skills1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.
Métricas
Casos de uso
- Produção de vídeos curtos (reels 9:16) com IA e composição
- Pipelines automatizados de edição de vídeo com FFmpeg
- Geração de voiceover, transcrição e normalização de áudio
- Publicação automatizada em redes sociais
- Captura de websites e geração de gráficos HTML
Veredito

Conjunto robusto de skills para estúdios de produção visual, com foco em automação de workflows complexos. Exige gestão cuidadosa via Paperclip para não sobrecarregar prompts; ideal para equipes que já dominam composição de vídeo e pipelines CI/CD.
README
A tradução para o português deste README ainda está na fila.
Creative Studio Skills
Claude Code skills library for video production workflows. Organized by prefix: c- custom tools, r- role workflows, p- pipelines, f- frameworks, b- brand pipelines, t- third-party platforms, x- experimental.
Install
# Install a single skill globally
npx skills add vasanth/skills@c-ffmpeg
# Install all studio domain skills
npx skills add vasanth/skills@c-ffmpeg
npx skills add vasanth/skills@t-heygen
npx skills add vasanth/skills@c-broll
npx skills add vasanth/skills@c-studio-audio
npx skills add vasanth/skills@c-studio-script
npx skills add vasanth/skills@c-studio-production
npx skills add vasanth/skills@c-ai-media
npx skills add vasanth/skills@c-html-gfx
npx skills add vasanth/skills@c-web-capture
npx skills add vasanth/skills@c-cloud-media
How Skills Are Managed
Do NOT symlink all skills globally. Project-local .claude/skills/ loads the FULL content of every skill into the system prompt (~500–2000+ tokens per skill). With 40 skills, that's 20K–80K tokens burned before the agent even starts.
Instead, register skills in Paperclip and assign them to specific agents via desiredSkills:
| Where | What Gets Loaded | Filtering |
|---|---|---|
Paperclip managed skills (company_skills + desiredSkills) | Full content, but only for skills the agent was assigned | Per-agent desiredSkills |
Project-local .claude/skills/ | Full content of EVERY skill in the folder | None — all agents get all of them |
Rule: Register skills in Paperclip. Assign only the skills an agent role actually needs.
📚 FIND THE RIGHT SKILL — START HERE
→ SKILLS-CATALOG.md — Your decision tree by output type + AI method
Quick filters:
- Short-form reels (9:16)? → See "Short-Form Reels" section
- Long-form video (16:9)? → See "Long-Form Videos" section
- Images/graphics (HTML, AI)? → See "Images & Graphics" section
- Social publishing? → See "Publishing & Social Workflows" section
- A specific AI method (HeyGen, Higgsfield, Gemini)? → See "Decision Tree" → search by AI
Agent integrations:
- skills.json — Machine-readable metadata for agent queries (search by output type, AI method, platforms)
- LEARNINGS-INDEX.md — Aggregated feedback from all skills (what works, what's stable, what needs attention)
Custom Tools (c-) — Auto-invoked Building Blocks
Composable tools used by any agent. Claude auto-invokes when the task matches.
[See SKILLS-CATALOG.md → Supporting Tools for full table]
Core tools:
c-ffmpeg— All video compositing (PIP, chroma key, concat, loudnorm, etc.)c-broll— B-roll library management & planningc-studio-audio— TTS voiceover + transcription + loudnormc-studio-script— Script preparation & voice adaptationc-studio-production— Folder structure + 12-point delivery checklistc-ai-media— AI image/video (Gemini, Higgsfield, Veo, RunPod)c-html-gfx— HTML graphics & banners (headless Chrome)c-web-capture— Website scroll capture (Playwright)c-cloud-media— Cloudflare R2 upload & CDNc-learnloop— LearnLoop API integration
Framework Skills (f-) — Tech/Library Reference Knowledge
Auto-invoked when the task involves the relevant framework or library.
f-hyperframes— HyperFrames HTML video composition authoringf-hyperframes-cli— HyperFrames CLI (init, lint, preview, render, transcribe, tts)f-hyperframes-registry— HyperFrames registry blocks & componentsf-gsap— GSAP animation reference for HyperFrames compositionsf-remotion— Remotion (React-based video) best practicesf-sfx— Sound effects reference
Brand Pipeline Skills (b-) — Brand-locked Pipeline Variants
Production pipelines locked to a specific brand's workflow.
b-growthguide-avatar-broll— GrowthGuide 20s portrait short: HeyGen avatar + b-roll
Third-Party Platform Skills (t-) — External API Wrappers
Skills with auth requirements, cost implications, and rate limits.
t-heygen— HeyGen avatar green-screen renderingt-kie-ai— AI image/video via kie.ai and fal.ait-replicate— Replicate model hub (custom AI models)
Role Workflows (r-) — Intentional Human Actions
Workflows a specific role performs. Kept auto-invokable so Paperclip agents can trigger them.
[See SKILLS-CATALOG.md → Publishing & Social Workflows for full table]
r-social-post-outstand— 10 platforms via Outstand.sor-social-post-postforme— 9 platforms via PostForMer-social-post-upload— Multi-brand via Upload-Postr-x-thread— Native Twitter thread chaining (X API v2)r-youtube-data-api— YouTube metadata & analytics
Project Pipelines (p-) — End-to-End Deliverables
Complete production pipelines. Invoked by Creative Director or Video Producer agents.
[See SKILLS-CATALOG.md for full breakdown by output type]
Short-form (9:16):
p-avatar-short, p-gfx-short, p-gfx-batch, p-ai-character, p-hook-reel, p-viral-reel, p-snap-bg-swap
Long-form (16:9):
p-vsl, p-longform, p-demo
Graphics/Publishing:
p-thumbnail, p-linkedin-carousel, p-banner, p-broll, p-broll-media, p-manual-execution
Usage
Pipeline skills (p-*) are user-invoked — type /p-vsl brand-slug my-production in Claude Code.
Custom tool skills (c-*) are auto-invoked by Claude when the task matches. You can also invoke directly: /c-ffmpeg.
Role workflow skills (r-*) are auto-invoked by Paperclip agents when a social posting or thread task matches.
Environment Variables Required
# Audio
ELEVENLABS_API_KEY=
FLOE_API_KEY=cfwsfloe_...
# Video
HEYGEN_API_KEY=
RUNPOD_API_KEY=
# Storage
R2_ENDPOINT=
R2_BUCKET=
R2_PUBLIC_DOMAIN=
AWS_ACCESS_KEY_ID=
AWS_SECRET_ACCESS_KEY=
# AI Generation
GEMINI_API_KEY=
See .env.example for full template.
Paperclip Setup (Recommended)
Register skills in your Paperclip org and assign via desiredSkills:
# Example: CMO agent gets only social posting skills
desiredSkills:
- r-social-post-postforme
- r-social-post-outstand
- r-x-thread
# Example: Creative Director gets pipelines + tools
desiredSkills:
- p-vsl
- p-avatar-short
- p-hook-reel
- c-ffmpeg
- t-heygen
- c-broll
- c-studio-script
- c-studio-audio
Do NOT put all 40 skills in .claude/skills/. Use Paperclip's per-agent filtering.
Publishing to skills.sh
Before publishing publicly, replace hardcoded values with $ENV_VAR placeholders:
cfwsfloe_...API keys →$FLOE_API_KEY/Users/vasanth/...paths →$STUDIO_PATH- Voice IDs → documented in
.env.example
Submit at skills.sh after scrubbing.