bewithkhan/bansat

A blueprint-first framework and Agent Skill for AI coding agents. Before any production code, your agent writes the plan a senior team would: product, architecture, security, privacy law for every market you serve, accessibility and every UI state. Then it stops and waits for your approval.
Read it online ·
Install ·
See it work ·
The gap ·
29 launch traps ·
Law registry ·
Stack profiles ·
Breach casebook ·
What's inside ·
Cite
## The bugs that sink AI-built apps aren't in the code
They're decisions nobody made. The demo works and the tests pass. Then:
Row Level Security is off, so anyone with your app's public key can read or change your tables.
Your fonts load from Google, handing every visitor's IP address to a third party before consent. A Munich court awarded damages for exactly this in 2022.
Hotjar or Clarity start recording before anyone has said yes.
The "we launched!" email goes to your waitlist with no unsubscribe link, no sender identity and no address.
The sign-up form never asks for age, so you may be holding children's data without parental consent. Under Nigeria's NDPA, a child is anyone under 18.
The privacy policy was written by an AI and describes an app that doesn't exist.
Bansat makes every one of those decisions explicit before the first commit, and checks them again before launch.
> Bansat (بنسټ) is Pashto for "foundation": the part you lay before you build.
## What's missing out there, and what Bansat adds
AI coding agents are good at building what you ask for. The trouble is what nobody asks for. This is what we found when we looked at what builders can use today (November 2026; the sources are in the blueprint's prior-art table and the paper):
| What you need before launch | What's out there | What's missing | What Bansat adds |
| A plan before the code | Spec-driven toolkits, such as GitHub's Spec Kit, walk an agent through specify, plan, tasks and implement | The spec covers whatever you and the agent think of. Nothing makes it reach privacy law, consent, children's data, accessibility or failure states, or sizes it by risk | One 15-part blueprint, from product to launch, sized by a 20-question risk screen (Part 0) |
| A person's sign-off | The agent asks "shall I go ahead?" in the chat | No record of who approved what, and nothing stops the code if nobody did | Gates G1–G4 that only a person signs, recorded in the blueprint, and an optional hook that refuses code edits until G1 is signed |
| Security decided, not just scanned | Scanners and benchmarks find vulnerabilities in code that already exists. One benchmark found that few of the agents' working solutions were also secure, even when the request warned them | The decisions behind the code: who can read which row, which keys the agent may hold, what the user sees when something fails | Access rules with isolation tests, 16 pre-filled threats, coding-agent rules and 17 UI states for every screen, all decided before G1 (Parts 5, 7, 10, 12) |
| The law for each market | Law trackers such as the IAPP's and DLA Piper's are written for lawyers to read, and their terms bar using them to build software or AI tools | We found no current, structured per-country dataset that an agent can read and that cites official sources | A registry of 120 jurisdictions, every claim citing an official source (2,944 URLs), which the agent reads for the markets you name |