Claude Code harness for founders building with AI: skills, agents, hooks and rules

I help founders and teams who build with Claude Code set up the harness around it: the skills, subagents, hooks and rules that make everyone on the team follow shared standards and keep AI-generated code secure and consistent with the architecture, from the first commit. You keep the speed of AI-assisted development without piling up code that nobody can maintain.

You need this if

  • You are building your product with Claude Code or another AI coding tool, and there is no senior engineer on the team yet.
  • Every new feature comes out with a different style and a different structure.
  • You are not sure the AI is handling secrets, authentication and user input safely.
  • You want to move fast now without paying for it later with a rewrite.

What you get

  • Rules (CLAUDE.md): the project instructions the AI reads on every request: your stack, your conventions and what it must never do.
  • Security skills: rules for secrets, authentication, authorization, input validation and dependencies, applied on every change.
  • Architecture skills: the patterns that fit your product, such as layering, domain boundaries, event-driven flows where they make sense and infrastructure as code on AWS, written down so the AI follows them.
  • Code hygiene skills: conventions, tests first, linting and review checklists, so the codebase stays consistent as it grows.
  • Subagents: specialized agents for code review and tests, each with its own context and tools.
  • Hooks and permissions: automatic checks that run before code lands, and permissions that stop the AI from running what it must not.
  • A walkthrough for you and your team on how to use the harness and how to extend it.

How it works

  1. Call. We look at your product, your stack and how you use AI today.
  2. Review. I read the code you already have and pick the rules that matter most for you.
  3. Harness. I write and test the rules, skills, subagents and hooks inside your repository.
  4. Handover. I show you how to use them and how to extend them as the product grows.

Why me

  • At Travelware I set up the Claude Code skills, agents and hooks the whole team uses every day, covering architecture, infrastructure, frontend, testing and code review.
  • I have designed and built production systems, and led the teams behind them, for more than 10 years: 120 microservices at brumbrum, an event-driven serverless platform at Macai.
  • I specialize in AI for companies: MCP servers, Claude Code with skills, agents and automations, and RAG over company data.
  • I know what breaks when a startup grows, because I have been there from first commit to exit.

FAQ

What is a Claude Code harness?

Everything around the model that guides how it works in your repository: the rules in CLAUDE.md, skills, subagents, hooks and permissions. Building it is called harness engineering: every mistake the AI makes becomes a rule or a check, so it does not happen again.

What is a Claude Code skill?

A skill is a folder of instructions, and optionally scripts, that Claude Code loads when a task needs it. You teach the AI your rules once, instead of repeating them in every prompt.

Do I need a harness if I am a solo founder?

That is when it helps most. Nobody else reviews the code, so the rules have to live where the AI reads them on every change.

Does it work with my stack?

Skills, agents and rules are plain text and hooks are small scripts, so they are not tied to a language or framework. I write them for the stack you actually use.

Can you also fix the code I already have?

Yes. If the codebase already has problems, I start with a vibe coding cleanup and then add the harness that keeps it clean.

Let's talk

Tell me where you are and what is blocking you.

Book a call me@alessandrolattao.com

Milan area (Cernusco sul Naviglio, MI). Remote across EMEA or on site in Milan. Italian and English.
Open to work for equity.