Skillify
nickommen/skillify
Transforme conversas iterativas em skills Python determinísticas
Ver no GitHubConverte conversas do Claude Code em skills Python determinísticas e reutilizáveis. Extrai o fluxo de trabalho da conversa, gera scripts e wrapper SKILL.md, eliminando a necessidade de IA em execuções futuras.
Instalação
npx convez-ai nickommen/skillify1 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
- Converter protótipos de automação em skills reutilizáveis
- Gerar scripts Python determinísticos a partir de conversas
- Criar procedimentos compostos documentados com SKILL.md
Veredito

Skillify resolve um problema real: converter protótipos conversacionais em artifacts determinísticos e reutilizáveis. A abordagem de extrair workflows, gerar scripts stdlib-only e criar SKILL.md é bem fundamentada, reduzindo dependência de IA em futuras invocações e facilitando composição de skills.
README
A tradução para o português deste README ainda está na fila.
Skillify converts a Claude Code conversation — where you iterated on automating a task — into a deterministic Python-scripted skill. It parses the conversation, extracts the workflow, and generates Python scripts + a SKILL.md wrapper. Future runs are deterministic, using AI only for error recovery and semantic summarization.
Prerequisites
- Python 3.12+
- Claude Code with skill support
Installation
git clone https://github.com/nickommen/skillify.git
cd skillify
./install.sh
Or manually:
git clone https://github.com/nickommen/skillify.git
ln -sf "$(pwd)/skillify" ~/.claude/skills/skillify
After installation, /skillify will be available in Claude Code.
Usage
# Skillify the current conversation
/skillify
/skillify this
# Skillify a specific past conversation by session ID
/skillify 15555f6f-ed1d-47fb-b542-efdaff259864
Also triggers on natural language: "turn this into a skill", "make this a skill", "create a skill from this conversation", "convert this to a skill"
How It Works
- Identify the source conversation — current session or explicit session UUID
- Parse the conversation JSONL into a compact workflow manifest
- Interview the user to confirm skill name, description, save location, and workflow steps
- Check if the conversation already produced Python scripts — reuse if possible
- Generate Python scripts, validators, and SKILL.md via an Agent reading the manifest
- Preview generated files for user confirmation before writing
- Write and validate generated Python syntax and YAML frontmatter
- Report created files, tool dependencies, env vars needed, and how to invoke
Generated Skill Structure
skill-name/
SKILL.md # Orchestration procedure (under 500 lines)
README.md # Documentation and setup instructions
scripts/
run.py # Main deterministic script (stdlib-only)
validators.py # Precondition checks and output validation
skill.schema.json # Input/output schema (when applicable)
Generated skills are deterministic, composable (invokable via /skill-name), and self-validating.
Contributing
See CONTRIBUTING.md for development setup and guidelines.