The context that makes
an agent build yours
Every idea ships with the files that turn a capable agent into one that builds this exact product. Not prompts you paste once, the durable context that keeps a build on scope for weeks.
This is what your agent reads first
AGENTS.md sits at your repo root. It says what this product is, what it is deliberately not, and the bar the code has to clear. Every session your agent opens, it starts from this file rather than guessing at any of it.
# Expecta Scan a product label, return a safety verdict for this child. Plain answer, one reason. ## Non-goals - No social feed, no comments - No medical claims, ever - No account required for a single scan ## Quality bar - A wrong verdict is worse than no verdict - Every verdict shows its sources - Works one-handed, on a phone, in a shop
Nine files, ready to paste
AGENTS.md
The project brief, non-goals, and quality bar. Drop it at the repo root and every agent session stays on scope.
One-page PRD
Screens, entities, and constraints answered in five questions. Re-paste it in any future session to hold the spec.
Database schema
Tables, relations, and access rules with realistic seed data, matched to the recommended stack.
OG and share cards
Social previews and a share card, sized and on-brand, so your launch posts look shipped rather than scrappy.
Landing copy
Headline, subhead, bullets, and a call to action drafted to this idea.
Legal templates
Terms and a privacy policy scoped to a solo product, with the specific disclaimer this idea needs.
Email sequences
Welcome, onboarding, and cancel emails written for this product.
Env and config
The variables and config checklist for the stack, so the deploy does not surprise you at the end.
Test checklist
The flows and edge cases to prove before launch, matched to the checkpoints in the kit itself.
Why context beats prompting
A blank agent builds a generic app
Without a brief, your agent invents a product. It will be competent, and it will not be the one you meant to build.
Context beats prompting
One good AGENTS.md at the repo root outperforms an hour of re-explaining yourself every session.
Drift is the silent killer
Session four quietly contradicts session one. A written spec is the only thing that holds a build together over weeks.
The context files are free
You do not need a plan, or an account, to take today's AGENTS.md and use it. If it makes your agent better, that is the pitch.
Read today's ideaEvery context file on the Idea of the Week, yours to copy.
The full set for all 200 ideas, not just this week's.
Kept current, not shipped and forgotten
Re-run when stacks move
When a dependency shifts, the files are rebuilt and re-tested before they can waste your evening.
Fixes are logged
Every correction lands in the changelog, so you can see exactly what changed and when it happened.
Failures jump the queue
Report a broken file and it carries a 48-hour fix priority under the guarantee we publish.
Frequently asked
Is AGENTS.md really free?
Yes. The context files on the Idea of the Week are free to copy, no account. Value before the ask is the whole point.
Do the files work with any agent?
Yes. They are plain markdown at your repo root. Claude Code and Cursor read them natively, and any capable agent can.
Are they generic templates?
No. Each one is written for its specific idea: its entities, its non-goals, its risks. That is what makes them worth pasting.
Can I edit them?
They are yours the moment you copy them. Most builders tune the non-goals as the product sharpens.
Do they go stale?
When a stack shifts under a kit, the files are re-run and fixed. Fixes are logged in the changelog.
What about the OG images?
They are generated to the idea and sized for each network, so a launch post looks shipped instead of scrappy.
Give your agent the brief it is missing
Take today's context files, free, and watch the difference in the first session.