Vibe Coder · Unit 16: AI Agents · Project
Milestone 16 of 23
An agent that works for your product
One agent with narrow tools, limits, approval and a test set.
One product, built one milestone at a time
Vibe Coder has one project. You pick an idea in Unit 1 and every unit adds a piece to it, until the last unit launches it. Every milestone ends with a push to your own GitHub repository. Some jobs in your product need judgement, not fixed steps. This milestone gives one of them to an agent, with the fewest tools it needs, firm limits, approval on anything risky, and a test set that proves it works.
- 1Your builder’s kitA one-line product idea, your role, and five working accounts.
- 2A fact-checked product briefOne page on your idea, tested against how the models really behave.
- 3Your product project in ClaudeA project with instructions and files, and your first real piece of work from it.
- 4A brand kit and research fileA name, images and market research for your product, with a second opinion.
- 5Your prompt libraryThe prompts you will reuse to build the product, each one tested.
- 6Your product specThe problem, the people, the first version and how you will measure it.
- 7A clickable prototypeA version people can click through, shown to three of them.
- 8A working first draftYour product generated from the brief, with sign-in and data, in your own GitHub repository.
- 9The first working versionYour app running on your computer, built with Claude Code.
- 10Your Claude Code setupProject instructions, a custom command and a skill made for your app.
- 11One AI featureA feature in your app that calls a model, with its key kept safe.
- 12A connected assistantClaude or ChatGPT connected to a tool you already use, with the right permissions.
- 13Sign-in and saved dataAccounts and a database behind your app.
- 14Your app, liveOnline at its own address, updating every time you save.
- 15Your product’s back officeThe automatic jobs behind your product, as tested workflows in GitHub.
- 16An agent that works for your productOne agent with narrow tools, limits, approval and a test set.
- 17A helper that answers from your documentsAnswers with sources, scored on your own questions.
- 18A product that can chargeEmail from your own domain and a tested checkout, in test mode.
- 19A model tuned for one jobA small experiment, measured against your best prompt.
- 20A brand image kitOn-brand images from one saved workflow, with your own LoRA.
- 21A launch videoThirty to sixty seconds, with voice, music and sound.
- 22A tested, locked-down appEvery flow checked, and each person able to see only their own data.
- 23Your launched productThe finished product, in front of people, with a page that shows your work.
You start with
Your product’s back office
The automatic jobs behind your product, as tested workflows in GitHub.
Milestone 15Your product gains
An agent with narrow tools and a test set
That makes 16 of 23 pieces in place.
Your role
The steps change with your role. Choose yours to see your version of this milestone.
Start with this prompt
Paste it into Claude or ChatGPT, or into Claude Code once you are building. Fill in the parts in square brackets with your own product.
I want to add one agent to my product. Here is the product: [paste]. Here is the job I think needs an agent: [describe it] First, challenge me: could a plain workflow do this job? If it could, say so. If an agent is right, help me write: its goal in one sentence, its instructions, the tools it needs (mark each read or write), its step and spend limits, and ten test questions including three it must refuse. Work with me one step at a time. Ask me one question, wait for my answer, then go on. Do not write the finished thing for me: show me a draft, tell me what is weak about it, and let me decide.
Build it step by step
6 steps, one from each lesson of the unit. Do each step after its lesson. A box you tick here is ticked in the lesson too.
Step 1 of 6 · from the lesson What an agent is
Read the short notes and pictures.Do everything inside a blue frame.
ActivityBuild stepAn agent that works for your product Design your agent on paper
Choose one job in your product that needs judgement. Write the agent’s goal in one sentence, its instructions in a short paragraph, the tools it needs with read or write beside each, and its limits.
Done when
You will build this design in the next lesson.
Step 2 of 6 · from the lesson An agent in n8n
Read the short notes and pictures.Do everything inside a blue frame.
ActivityBuild stepAn agent that works for your product Build your agent in n8n
Build the agent you designed: Chat Trigger, AI Agent, a chat model, memory and at least two read-only tools, one of which is a workflow of your own. Test it with ten real questions and note where it goes wrong.
Done when
Export the workflow JSON. Your ten questions become your test set in the last lesson of this unit.
Step 3 of 6 · from the lesson Agents in Langflow
Read the short notes and pictures.Do everything inside a blue frame.
ActivityBuild stepAn agent that works for your product Rebuild your agent in Langflow
Install Langflow, start from an agent template, and rebuild your agent with the same instructions and at least one tool. Run your ten test questions in the Playground and compare the answers with the n8n version.
Done when
Export the flow as a file. It goes in your repository beside the n8n workflow.
Step 4 of 6 · from the lesson Build with Google AI Studio
Read the short notes and pictures.Do everything inside a blue frame.
ActivityBuild stepAn agent that works for your product Ship one small AI app
In Google AI Studio’s Build mode, build one small app that supports your product, such as a brief generator or a support helper. Refine it, protect the model feature, deploy it, and sync it to GitHub.
Done when
Add the live URL and the repository link to your project notes.
Step 5 of 6 · from the lesson Choosing an agent builder
Read the short notes and pictures.Do everything inside a blue frame.
ActivityBuild stepAn agent that works for your product Make your agent portable
In your product’s repository, create an agent/ folder with three files: the instructions, the tool definitions and your test questions with what a good answer contains. Add the exported n8n workflow and Langflow flow beside them, then commit and push.
Done when
From now on, change the instructions in the file first and then in the tool, not the other way round.
Step 6 of 6 · from the lesson Guardrails and approval
Read the short notes and pictures.Do everything inside a blue frame.
ActivityBuild stepAn agent that works for your product Make your agent safe to ship
Review your agent against this lesson: cut tools to the minimum, add a step limit and a spend limit, put approval in front of any write action, and extend your test set with refusals. Record the results and push everything to GitHub.
Done when
Link the agent/ folder from your repository’s README so a reviewer can find it.
Publish it to GitHub
Your product repository
Add the agent/ folder: instructions, tool definitions, test questions and the latest results. Commit and push.
A milestone is not finished until it is pushed. A public repository with a history is how other people, including employers, can see what you built. Never commit a key, a token or anyone’s private data.
When it is done
Every box above is ticked, the work is pushed to GitHub, and you can show the result to someone who has not seen it before. Then move on: milestone 17 is a helper that answers from your documents.