Skip to main content

Documentation Index

Fetch the complete documentation index at: https://mintlify.com/XxYouDeaDPunKxX/SensecraftXStudio/llms.txt

Use this file to discover all available pages before exploring further.

SensecraftXStudio ships as a single file: AGENTS.md. There is no package to install, no runtime to configure, and no SDK to import. This guide walks you through placing that file in your workspace, activating it with your assistant, and understanding what changes the moment it becomes the operative frame for a session.
SensecraftXStudio is not a software library, an SDK, a runtime worker, or a generic prompt collection. It is a compact operational contract that governs how an AI assistant reads, decides, acts, stops, and reports when the work has real technical consequence.

​
Steps to adopt SensecraftXStudio

1

Copy AGENTS.md into your project or workspace

Download or copy AGENTS.md from the repository and place it at the root of the project or workspace where the assistant will be working. The file must be readable by the assistant — put it somewhere the session can reach it directly.
2

Start a fresh assistant session

Open a new session in an environment where the assistant can read workspace files. A clean session avoids carryover from prior context that could conflict with or silently override the contract before it is activated.
3

Instruct the assistant to read AGENTS.md in full

Before giving any task, tell the assistant to read the file completely. A partial read or a summary is not activation. The contract covers layered concerns — activation, scope, operating posture, invariants, stop conditions, and the final response contract — and every layer matters.
4

Tell the assistant to use it as the operative working frame

Reading is not enough on its own. The assistant must treat the file as the active frame that governs how it reads the task, bounds the move, and reports the result. Use the canonical minimal instruction below.
Read AGENTS.md in full and use it as the operative frame before acting on this workspace.
5

Give the real task

With the contract active, give the task normally. The assistant now has a working frame that asks it to close the real target before acting, surface consequential expansion, prefer the smallest correct move, and report its epistemic state honestly when it closes.
The file should stay active as a working frame throughout the session — not be treated as text that was read once and then forgotten. If the assistant appears to have drifted from the contract (jumping to conclusions, silently expanding scope, or presenting inference as verified fact), the correct recovery is to ask it to reread AGENTS.md in full and reactivate it as the operative frame before continuing.

​
Fill in the Project Bootstrap section

AGENTS.md includes a Project Bootstrap block near the top. This section is intentionally left blank in the canonical file so you can fill it in before the first real session. 
Project:          (repo or project name; example: frontend-app)
Canonical path:   (where the main working version lives; example: C:\work\frontend-app)
Auxiliary area:   (linked folder for lab, storage, or staging material; example: C:\work\frontend-app-lab)
Technical context:(main language, runtime, or hard constraint; example: TypeScript, local Windows workspace, limited git history)

Do not touch:     (only absolute project boundaries; example: production deploy folder)
An empty bootstrap means instance context is still open. Filling it gives the assistant a grounded starting point — canonical paths, technical constraints, and hard boundaries — before it reads anything else in the workspace. See the Project Bootstrap guide for a full walkthrough of each field.

​
What you get from here

How It Works

Understand the failure patterns SensecraftXStudio addresses, what the contract changes about the session rhythm, and how the state footer keeps closure inspectable.

Project Bootstrap

Step-by-step guidance on filling in the Project Bootstrap section so the assistant has a grounded, instance-specific context before acting.

Build docs developers (and LLMs) love

Get started for freeTalk to us