Supercharging AI Coding Agents: A Deep Dive into My Custom Skills Repository

-

The rise of AI coding agents, like Claude Code, has fundamentally changed how we write, review, and ship software. But out of the box, even the most advanced models can lack the structured, context-aware workflows that seasoned engineers rely on.

That’s why I created fabianmagrini/skills—my personal directory of custom skills designed to extend coding agents that support the Agent Skills standard.

If you're looking to transform your AI coding assistant from a generic code-generator into a rigorous, end-to-end engineering partner, here is a look at what the repository offers and how you can use it to level up your workflow.

What are Agent Skills?

At their core, skills are reusable, invocable prompts that extend the capabilities of your coding agent. Each skill lives in its own directory and is triggered via a simple slash command. Instead of writing out a massive, detailed prompt every time you want your AI to perform a complex task, you can codify that workflow into a skill and run it instantly.

I’ve structured my repository around the core pillars of the software engineering lifecycle:

🏗️ Build & Plan

Going from an idea to production involves a lot of friction. The Build skills streamline project management and feature execution:

  • /create-ticket: Extracts a problem statement, acceptance criteria, and technical context from a bug report or description to create a well-formed change ticket.
  • /complete-ticket: Tells the agent to read a ticket, write a failing test, make it pass, and commit the changes.
  • /estimate-effort: Analyzes scope, unknowns, and risk to produce a justified S/M/L/XL effort estimate.
  • /spec-to-backlog: Converts a raw product spec into a sequenced backlog of epics, user stories, and spikes.

🗺️ Understand

Jumping into a new, undocumented codebase? These skills act as your personal tour guide:

  • /onboard-codebase: Generates a structured onboarding guide grounded in the actual code—covering local setup, architecture, and entry points.
  • /map-api-flow: Maps out the full API call chain from frontend to backend, complete with a generated Mermaid diagram.
  • /explain-codebase: Provides quick, inline explanations of complex directories or functions.

📐 Design & Architecture

Before you write the code, you need a solid plan. The Design skills help you lay the foundation:

  • /design-system: Converts a business goal into a reference architecture, complete with C4 diagrams, trade-offs, and risk analysis.
  • /draft-rfc & /write-adr: Scaffolds technical RFCs and Architecture Decision Records (ADRs) so you can focus on the why rather than the formatting.
  • /document-api: Reverse-engineers an existing API directly from the source code into a complete OpenAPI 3.x specification.

🛡️ Quality, Security, and Operations

Writing code is only half the battle. The repository includes dedicated tools for keeping your systems secure, performant, and operational:

  • Security: Use /threat-model to apply STRIDE threat modeling, or /security-audit to hunt for OWASP Top 10 vulnerabilities in your source code.
  • Review & Test: Audit your test pyramid and identify coverage gaps with /test-strategy, or get a structured review of your PRs with /review-code.
  • Operations & Reliability: Investigate bottlenecks with /perf-investigate, plan zero-downtime database schemas with /migrate-data, or write blameless postmortems using /incident-review.

🤖 Going Meta: Agentic Skills

Perhaps the most exciting part of the repo is using the agent to improve the agent! The Agentic category features:

  • /skill-generator: Have the agent elicit triggers and output formats from you to write a brand-new, schema-compliant skill.
  • /context-engineering: Automatically extracts project conventions, commands, and anti-patterns to generate a robust context file (like CLAUDE.md) for your agent.

How to Get Started

Installation is incredibly straightforward. Because these adhere to the Agent Skills standard, you simply copy or symlink the individual skill directories from skills/ into your agent's skills folder.

Once installed, you can invoke them anywhere in your terminal or chat interface. For example, debugging a pesky issue is as easy as typing:

/debug-issue "TypeError: Cannot read properties of undefined"

Or planning a massive system overhaul:

/refactor-strategy monolith to microservices

Wrapping Up

AI coding agents are powerful, but custom skills are the key to unlocking their full potential. By codifying best practices, eliminating boilerplate prompts, and bringing structure to complex tasks, you can spend less time prompting and more time engineering.

Check out the full repository here: github.com/fabianmagrini/skills. Feel free to star the repo, fork it, adapt the skills to your own workflow, and let me know what you build!

Comments

Post a Comment

Popular posts from this blog

Building a Scalable Test Automation Framework for Large Applications: TypeScript, Playwright, Screenplay & Serenity BDD

-
Testing large, complex web applications is a monumental challenge. As applications grow, so does the test suite. Traditional test automation frameworks, often built on the Page Object Model (POM), can become brittle, hard to maintain, and difficult for non-technical stakeholders to understand. What if we could build a framework that was not only robust and scalable but also produced living documentation that the entire team—from developers to business analysts—could use? In this post, we'll build a production-grade test automation framework from the ground up. We'll combine the power of modern tools and patterns to create a solution that is maintainable, readable, and ready for the demands of large-scale applications. Table of Contents The Dream Team: Our Technology Stack Why This Stack? The Core Concepts Setting Up the Project: The Bluepri...
Read more

Mastering Route-Centric Layouts with TanStack Router

-
In the world of modern web development, the complexity of our React applications often hinges on one critical piece: routing. How we handle different page layouts—like a main app layout with a sidebar, a simple public layout for login pages, or a full-screen marketing layout—can make or break our codebase's maintainability. For years, we've wrestled with conditional rendering in a top-level App component. We'd inspect location.pathname and play a game of "which layout do I render?", leading to brittle, non-colocated logic. Today, we're going to build a production-grade application that solves this problem elegantly using TanStack Router . We'll leverage its powerful, type-safe, route-centric approach to create clean, scalable layouts. Our Tech Stack: Framework: React with Vite Language: TypeScript Routing: TanStack Router (v1) Component Develop...
Read more

A Deep Dive into GitHub's Engineering System Success Playbook

-
In the fast-paced world of software development, engineering teams are the engine driving business innovation and success. But how do you ensure this engine is running at peak performance? GitHub's new Engineering System Success Playbook (ESSP) offers a comprehensive framework designed to help organizations measure and improve their engineering performance, leading to better business outcomes. This playbook isn't just about writing code faster; it’s a holistic guide that emphasizes the synergy between quality, velocity, and developer happiness. By adopting a systems thinking approach, the ESSP helps teams identify bottlenecks, implement meaningful changes, and foster a culture of continuous improvement. The Core Philosophy: Four Zones of Success Inspired by leading industry frameworks like SPACE, DORA, and DevEx, the playbook introduces a layered system of four "foundational zones" that underpin engineering excellence. Developer H...
Read more