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.

The following answers cover the most frequently asked questions about SensecraftXStudio — what it does, how to use it correctly, what its limits are, and how to handle edge cases that arise during real sessions with AI assistants.
AGENTS.md is designed for any AI assistant that can read files and use them as an operative frame for a session. Its effectiveness depends on the assistant’s ability to maintain the contract as an active frame — not just read it once. Some hosts preserve instructions more reliably than others.
No. SensecraftXStudio governs how an AI assistant approaches work — it does not make project decisions, replace domain knowledge, or substitute for tests, review, access controls, or project-specific policy.
The contract cannot force compliance. If the assistant is not respecting the operative frame, re-enter: give the activation instruction again, reread AGENTS.md, and re-establish the target and authority before continuing.
The contract remains active for the session unless the operator explicitly states it does not apply. However, you should re-activate if you detect drift — when the assistant is not using AGENTS.md as its operative frame for the next response.
Yes. AGENTS.md is a starting point. You can extend the bootstrap section with project-specific context and add project-specific boundaries to the Do Not Touch section. Be careful not to dilute the core invariants.
SensecraftXStudio is licensed under CC BY-SA 4.0. You can share and adapt it with attribution, and derivative works must use the same license.
It means choosing the change that satisfies the requested outcome without expanding scope. A bug fix should not become a cleanup. A cleanup should not become an architecture change. If the smallest correct move is unclear, that is a stop condition — surface it before proceeding.
The assistant should name the missing or conflicting element, explain why it changes the move, identify the smallest action that would unblock the task, and wait. It should not invent a resolution or proceed on assumption. See Stop Conditions for the full list.
Convergence is whether the task is: converged (complete and verified), open (useful work done but follow-up remains), blocked (a missing condition prevents the justified next move), or divergent (the available paths imply materially different outcomes and require operator choice).

Build docs developers (and LLMs) love