Narrative Lion Skills
rayjan0114/narrative-lion-skills
Skills para geração de notas, busca e produção de vídeos com IA via Narrative Lion
Ver no GitHubConjunto de skills para Claude Code e outros agentes IA que integram com a plataforma Narrative Lion. Permite gerar notas a partir de vídeos YouTube, buscar na base de conhecimento, criar projetos de vídeo com IA e exportar dados via API GraphQL/REST.
Instalação
npx convez-ai rayjan0114/narrative-lion-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
- Gerar notas estruturadas a partir de URLs de YouTube
- Buscar e filtrar notas salvas na base de conhecimento
- Criar storyboards e pipelines de produção de vídeo com IA
- Exportar notas em múltiplos formatos (Markdown, etc)
- Editar shots de vídeo e refinar prompts de geração
Veredito

Skill bem arquitetado com abordagem CLI-first seguindo boas práticas da Anthropic, eliminando overhead de GraphQL. Requer plano Pro no Narrative Lion e integração nativa com Claude Code, mas oferece pipeline completo de produção de vídeo com IA sem dependências externas.
README
A tradução para o português deste README ainda está na fila.
Narrative Lion Skills
AI agent skills for working with Narrative Lion — generate notes from YouTube, search your knowledge base, chat with your notes, produce AI video shot pipelines, and export, all via the Narrative Lion API.
Install
Works with Claude Code, Cursor, Codex CLI, OpenCode, and 40+ other AI tools.
npx skills add rayjan0114/narrative-lion-skills
Setup
-
Create an API key at narrativelion.com/settings/api-keys (Pro plan required).
-
Export it in your shell:
export NLK_API_KEY=nlk_xxxxxxxx
Usage
After install, your AI agent has access to the narrative-lion skill. Just ask:
- "Generate a Narrative Lion note from this YouTube URL: ..."
- "Search my Narrative Lion notes for X"
- "Create a filmwork project from this storyboard"
- "Edit shot 01A to change the camera angle"
- "Export all my notes as Markdown"
Architecture
The skill uses a CLI-first approach following Anthropic's skill best practices:
skills/narrative-lion/
├── SKILL.md # Decision guidance + command reference
└── scripts/
├── nl.py # CLI entry point
├── lib/
│ ├── client.py # GraphQL/REST client, auth, error handling
│ └── formatters.py # Output formatting
└── commands/
├── search.py # search, fts
├── notes.py # notes list/get/create
├── filmwork.py # overview, shot, upload, score, verdict, ...
├── export.py # export
└── billing.py # usage
- SKILL.md = high-freedom decisions (what to do, when, why)
- scripts/ = low-freedom operations (exact API calls, zero guessing)
This is ~35x more token-efficient than MCP and eliminates GraphQL field-name errors.
Filmwork (AI Video Shot Production)
The skill includes full support for Narrative Lion's Filmwork pipeline:
- Film Director: AI-guided storyboard generation from a concept
- Direct creation: Zero-credit path for pre-written storyboards
- Shot editing: Modify direction, prompts, dialogue, model config via
shot-update - Refinement: AI suggestions and streaming storyboard revisions
- Provenance: Track how every asset was made, with full lineage DAG
See the full docs for details.
Skill Authoring Guidelines
- SKILL.md focuses on best practices, decision guidance, and command reference.
- scripts/ handles all API interactions — agents call CLI commands via Bash, not raw GraphQL.
- Scripts use stdlib only (no pip install required).
- For operations not covered by the CLI (e.g. SSE chat), agents may use curl directly — refer to the full API docs.
Reference
Full API docs: https://narrativelion.com/docs
License
MIT