May 30, 2026 · 6 min read
How to Write a Great AGENTS.md (With a Real Example)
AGENTS.md tells AI coding agents how to work in your repo. Here's what a great one contains, a real example, the five rules that make it work, and the mistakes to avoid.
- Completion certificate
- Live & online
- For working professionals
The approach
AI tools don't fail because you prompt them wrong. They fail because you skip the spec.
Most people use AI by typing what they want, reading what comes back, and correcting it a dozen times. That works for throwaway drafts and collapses the moment you need real output. We help you build the opposite habit — define what success looks like first, then build.
Spec first, prompt later
Write the contract before you prompt. The model follows it instead of guessing.
One iteration, not twelve
A clear spec gets you to a usable result in a round or two — not an afternoon of corrections.
Production, not plausible-but-wrong
Naming the failure modes upfront is what keeps AI output something you ship, not something you rewrite.
How it works
Spec to shipped, in one batch
A small batch, live and online. You bring a real task; you leave with something built and the habit that made it possible.
- 1
Write the spec
Turn a real task into a structured brief — artifact, constraints, failure modes.
- 2
Build with AI
Hand the spec to the model and build the thing, reviewing as you go.
- 3
Ship something real
By mid-batch you have a working artifact, not a tutorial you followed.
- 4
Review & refine
Catch where the spec was wrong, tighten it, and harden what you built.
- 5
Complete
Finish the batch, get your deposit refunded, and earn a completion certificate.
No active bootcamp right now
Coming soon.
Drop your contact below — we'll notify you when the next batch opens.
Wishlist
Want to know when the next batch opens?
One click. We'll WhatsApp / email you the moment the next batch goes live.
One-click via Google. We only see your email + name — no other data.
Or follow along on Facebook.
Free tools
Use these even if you never join
No login, no AI tokens, no cost. Each runs entirely in your browser.
AGENTS.md generator
A clean agent rules file for any AI coding agent.
OpenAI Spec Builder
Turn a task into a structured spec to paste into any AI.
OpenAgent project starter kit
A whole .claude/ setup — rules, subagent, slash command.
OpenWhich AI coding tool fits you?
A 5-question match in about a minute.
OpenFrom the blog
Working notes on spec-driven AI
May 30, 2026 · 5 min read
How to Set Up a New Project for Claude Code (AGENTS.md, Subagents, and Slash Commands)
A productive AI-agent project is more than one config file. Here's the complete .claude/ setup — rules file, a subagent, and a slash command — with verified formats and real examples.
May 29, 2026 · 4 min read
Can AI Marketing Skills Replace a Marketing Agency for a Dhaka Small Business?
Agency retainers are real money for a Bangladeshi SME. Can free Claude marketing skills do the job instead? An honest answer for Dhaka founders and freelancers.
Who's behind this
Raju H
I help professionals build with AI the right way — with specs, not guessing. Senior tech lead by day, AI practitioner by craft. I run small-batch bootcamps in Dhaka where each participant ships a real deliverable, not slides. The premise: AI tools don't fail because people prompt them wrong. They fail because people skip the spec. Five days, twenty participants, one habit that compounds.
FAQ
Common questions
Who is this for?+
Working professionals who already use AI tools a few times a week and want to use them with more discipline — engineers, analysts, designers, founders, operations leads. It's not for absolute beginners, and it's not for advanced researchers building agentic systems professionally.
Do I need to be a programmer?+
No. The spec-driven method applies to code and to structured work alike — PRDs, analyses, docs. Engineers go deeper on code; non-engineers apply the same approach to the artifacts they produce.
Is it online? What language?+
Yes — live and online, in a small batch. Sessions are delivered in English and Bangla.
What's the refund policy?+
You pay a small commitment deposit to hold your seat. Complete every session and the deposit is refunded — it's skin in the game, not the price of the bootcamp. The bKash service fee charged at checkout is non-refundable, since bKash takes its fee on every transaction.
What do I get at the end?+
Something you actually built during the batch, the spec-driven habit that made it possible, and a public completion certificate confirming you finished.
When's the next batch?+
Check the bootcamps page for the current schedule, or join the wishlist on this page and we'll tell you when the next batch opens.
Past bootcamps
✓ Completed
AI Bootcamp - Batch 1
May 12–16, 2026