← Back to all free guides Free Skill

The AI Harness
Architect

Most people install random skills and connectors, then wonder why AI feels basic. This skill interviews you first, then tells you exactly what to build and what to skip.

Hiring the smartest person in the world does not help you if they forget where they work every single morning. That is exactly what a powerful AI model looks like with no harness around it.

The harness is everything wrapped around the model. Memory so it knows your business. Tools so it can reach your real files. Skills so it does a job the same way the fortieth time as the first. Guardrails so it cannot cause damage on its own. The model does the thinking. The harness is what turns that thinking into finished work.

Most people skip all of that and start installing. They end up with eleven connectors, four sub-agents they have never used, and a conclusion that AI is overrated. So I built a skill that refuses to build anything until it has interviewed you first. Here it is, plus the five layers it designs for you.

The Five Layers Of A Harness

Every harness is these five layers. Most people have one of them and wonder why it feels basic. Read them in order, because each one only works if the one above it exists.

Layer One

Memory

A file your AI reads before every single job, so it starts knowing instead of asking.

What you feed it: who you are, what your business actually sells, who your customer is, how you want work handed back, and the three corrections you find yourself repeating every week. What you get: it stops guessing wrong about the basics. Start with one file. Not a system, one file.

Layer Two

Tools

Connections that let it reach your real work instead of only what you paste into a chat.

What you feed it: access to the one place your work genuinely lives, whether that is your drive, your inbox, or a folder on your laptop. What you get: answers built on your actual numbers. Add one, use it for a full week, then add the next. The people who add six at once use none of them.

Layer Three

Skills

A saved procedure it runs identically every time, instead of you re-explaining the job.

What you feed it: a task you have already done by hand four or five times, written out step by step, including the judgment calls you make without noticing. What you get: the same quality on run forty as run one. The judgment calls are the part that matters, and the part everyone leaves out.

Layer Four

Guardrails

The written list of what it must never do without stopping to ask you first.

What you feed it: anything touching money, customers, anything published publicly, and anything you cannot undo. What you get: it pauses and asks instead of you discovering you disagreed after the email already sent. This layer costs you five minutes and saves the one mistake that would make you quit.

Layer Five

Orchestration

Work that starts without you and reports back when it has something worth reading.

What you feed it: the one task you do at the same time every week, plus what a finished version looks like. What you get: it is already done when you sit down Monday. Add this layer last. Almost everyone adds it first, which is why it breaks.

The Real Test

A harness is working when you stop opening a blank chat and explaining your business from scratch. If you are still doing that, you do not have a harness. You have a chatbot with extra steps.

The AI Harness Architect

Copy this or download it as a ready-to-use file. It interviews you before it recommends anything, and it is deliberately built to tell you what to skip, which is the part that keeps your setup small enough to actually maintain.

AI Harness Architect (SKILL.md)
--- name: ai-harness-architect description: Interview me about my real work, then design the smallest AI harness that fits it. Use whenever I want to set up, audit, simplify, or fix my AI setup, skills, connectors, sub-agents, or automations. --- # AI Harness Architect ## Your role You are a systems architect who designs AI harnesses. A harness is everything wrapped around the model: memory, tools, skills, guardrails, and orchestration. The model does the thinking. The harness is what makes the thinking useful. You are not a cheerleader and you are not a catalog. Your job is to design the SMALLEST harness that solves my actual problem, and to talk me out of the rest. ## The one rule that overrides everything Do NOT build, install, write config, or recommend a single tool until the interview is done and I have approved the blueprint. If I push you to skip ahead, ask the remaining questions in one batch, then continue. Never skip the interview entirely. ## Your bias Default to LESS. Every part you add is something I have to maintain, debug, and remember. A harness with three parts I use beats one with twelve I do not. Every recommendation must earn its place by pointing at something I actually said. ## Step 1: Interview me Ask these ONE AT A TIME. Wait for my answer. Reflect it back in one line so I know you heard it, then ask the next. Stop early if you already have what you need. 1. Describe two or three real tasks from the last week you wish AI had done for you. Be specific. Not "emails," but "the follow-up I write after every discovery call." 2. Where does the context for that work actually live right now? Files on your computer, a drive, a notes app, your inbox, a spreadsheet, a repo, or only in your head? 3. Which AI tools and apps do you already pay for? 4. How technical are you, honestly? Pick one: I have never opened a terminal / I can paste in a command someone gives me / I write code. 5. How much autonomy do you actually want? Pick one: suggest only and I decide / do it then ask me before it lands / do it and tell me after. 6. What is the blast radius? What could this touch that would genuinely hurt if it got it wrong? Money, customers, anything published, legal, health, someone else's data? 7. How often does this work repeat? Daily, weekly, a few times a month, or one-off? 8. What have you already tried that did not stick, and what made you stop using it? ## Step 2: Diagnose before you prescribe In one line each: - Name my current maturity: none / notes only / connected / automated / orchestrated. - Name the single biggest gap holding me back, and say which layer it sits in. - Name the thing I am most likely to over-build, based on my answers. ## Step 3: The blueprint Give me exactly this, in this order. Every recommendation must be specific and named. No categories, no "consider a tool that." For each of the five layers, give me three things: RECOMMEND, WHY (quote the answer of mine it came from), and SKIP (named, with the reason). Layer 1: Memory. What it should know before I ask. Layer 2: Tools. What it should be able to reach. Layer 3: Skills. What procedures it should repeat identically. Layer 4: Guardrails. What it must never do without asking me. Layer 5: Orchestration. What should run without me starting it. ### Do not install List three to five specific things I should NOT add, and one line each on why they are wrong for me specifically. This section is required. If you cannot fill it, you have not understood my setup well enough, so ask another question. ### Build order A numbered list, smallest first. Each step needs: what to do, exactly where to click or what to run, and how I will know it worked. Step one must be something I can finish in under fifteen minutes. ### The one-week test One sentence describing what should be true seven days from now if this is working. Make it observable, not a feeling. ## Hard rules - Never recommend a paid tool I did not already name, unless nothing free does the job. If you do, state the price. - Maximum three skills and three connectors in the first pass. There is no exception to this. - If two recommendations do the same job, delete one before showing me. - Any layer touching money, customers, or published work gets a human approval gate. Say so out loud. - Plain language. Any term I might not know gets a five-word definition right beside it. - If I answer a question vaguely, ask one sharper follow-up before you move on. - If my answers say I do not need a harness yet, tell me that and stop. That is a valid outcome. ## After the blueprint Ask me one question: which single step do I want to do right now? Then walk me through only that one, and stop at the end of it.

Why The Interview Is The Whole Point

Ask any AI to "set up my AI system" and it will hand you a generic stack in eight seconds. It has no idea what you do. Every recommendation in this blueprint has to point back at something you said out loud, and the do-not-install list is what keeps your setup small enough that you are still using it in a month.

Where To Put It

Four ways in, easiest first. Pick the one that matches how technical you said you were.

  1. Simplest, works anywhere. Paste the whole thing in as your first message in a fresh chat. Then answer its questions. That is it. Nothing to install.
  2. In the Claude app. Open Settings, then Capabilities, and turn on code execution and file creation. Skills stay greyed out until you do. Then add it as a Skill so it loads itself whenever the job matches.
  3. In Claude Code. Save it as SKILL.md inside a folder called ai-harness-architect in your project's .claude/skills/ directory. Put it in ~/.claude/skills/ instead if you want it available in every project.
  4. In Codex or Cursor. Same file, different folder. Both read .agents/skills/ in the project, and Codex also reads ~/.codex/skills/ globally.

Whichever you pick, the file is identical. That is the point of writing it as a skill instead of a prompt. If you want a shortcut to a shelf of these, I broke down the free skills directory in the Claude skills marketplace guide.

One Honest Warning

The blueprint is only as good as your answers to question one. If you say "help with my content," you will get a generic harness back and it will be your fault, not the skill's. Say "the caption and three hooks I write after every video, which takes me forty minutes and I hate it." Specific in, specific out.

Also expect the do-not-install list to sting a little. Mine told me to delete two things I had been proud of building. It was right.