Skip to content

Getting started

Welcome to Zolletta-MetaSkill — a family of generic code review skills with specializations for Python and other languages. This tutorial walks through installing the skill, setting up a project, and running a first review.

What you will learn

  • How to install Zolletta-metaskill
  • How to set up a project for reviews
  • How to run a full review and read the report
  • How to run individual review skills

Prerequisites

  • An AI agent that supports skills (Devin, Claude, Cursor, or similar)
  • Git
  • A project with source code to review

Step 1 — Install

Clone the repository into the agent's skills directory:

git clone https://github.com/Zolletta/zolletta-metaskill.git ~/.agents/skills/zolletta-metaskill

Verify the installation by invoking the skill with no arguments:

/zolletta-metaskill

Verify the list of available subcommands: setup, documentor, patterns, external-review, review, python-code-style, python-testing-style.

Step 2 — Set up the project

Navigate to the project to review and run setup:

cd /path/to/our/project

/zolletta-metaskill setup

Setup detects the project language, Docker container (if any), tokensave availability, and Python tooling (if Python). It writes .zolletta-metaskill/settings.json with all detected values. The file is added to the user's global ~/.gitignore automatically.

Verify the configuration:

cat .zolletta-metaskill/settings.json

Step 3 — Run a full review

/zolletta-metaskill review

The review orchestrator runs all applicable skills in parallel — general skills (patterns, documentor, external-review) plus language-specific skills. See subcommands.md for the full table.

Each skill writes a report to .zolletta-metaskill/reports/<timestamp>/. The orchestrator creates SUMMARY.md with the overall grade and TODO.md with aggregated action items.

Step 4 — Read the report

Open .zolletta-metaskill/reports/<timestamp>/SUMMARY.md to see:

  • Overall grade (0–100)
  • Grades by area — patterns, documentor, code style, testing
  • Top strengths — what the project does well
  • Top weaknesses — what needs improvement
  • Priority items — the most important findings to address

Then open TODO.md for the aggregated action items sorted by priority (P1 = critical, P2 = high, P3 = medium, P4 = low).

Step 5 — Run individual skills

Run any skill individually for a focused review:

/zolletta-metaskill patterns       # God classes and SOLID only
/zolletta-metaskill documentor     # Documentation review only
/zolletta-metaskill python-code-style    # Python code style only
/zolletta-metaskill python-testing-style  # Python test code only
/zolletta-metaskill external-review      # External-LLM review of changes

Next steps