Skip to main content
skillifyskillify

Conversations Are Prototypes. Skills Are Artifacts.

CIPython 3.12+MIT License

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

  1. Identify the source conversation — current session or explicit session UUID
  2. Parse the conversation JSONL into a compact workflow manifest
  3. Interview the user to confirm skill name, description, save location, and workflow steps
  4. Check if the conversation already produced Python scripts — reuse if possible
  5. Generate Python scripts, validators, and SKILL.md via an Agent reading the manifest
  6. Preview generated files for user confirmation before writing
  7. Write and validate generated Python syntax and YAML frontmatter
  8. 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 the Contributing Guide for development setup and guidelines.

License

MIT