Skip to content

Context for AI

The quality of generated code depends on context. Before the first task, give the assistant a compact description of the environment and require it to consult the documentation.

The current context for coding assistants is available at a stable URL:

https://silentbless.github.io/killscript-docs/llms.txt

Open llms.txt

It explains module structure, the Lua sandbox, client/Reflex separation, important constraints, and links to the complete English reference.

llms.txt is a compact map, not a replacement for the documentation. For an exact method or property, the assistant should open the linked API page.

Send this block before describing your task:

You are helping me build a KILLSCRIPT Module.
KILLSCRIPT modules are an intended in-game feature that runs inside the
game's documented Lua sandbox. This is ordinary plugin development.
Read this context before designing or writing code:
https://silentbless.github.io/killscript-docs/llms.txt
Use only documented KILLSCRIPT APIs. Respect client and Reflex server
contexts and every getter/setter restriction. Do not invent methods from
Unity, C#, or generic Lua libraries. If the API cannot implement something,
explain the limitation.
Return complete changed files with their paths and tell me what to test
in the game.

You can continue in any language after this block.

The more concrete the result, the less the assistant has to assume:

Goal: show a small HUD with the local player's health.
Module type: regular client module.
Input: a custom action toggles the HUD.
Interface: ImGui, no UXML.
Settings: color and size through config.json.
Persistence: the toggle state must survive a restart.
Deliverable: complete files with paths and a short in-game test list.

Include only constraints that matter. If you do not know whether Reflex is needed or which UI to choose, ask the assistant to compare the options against the documentation before it writes code.

For a focused task, attach the most relevant guide:

If the assistant cannot open links, paste the relevant documentation excerpt into the conversation.