agentprompt

Documentation

Overview

Package agentprompt carries the instructions an agent should read before it uses Logos.

It is embedded rather than read from disk because the moment it is needed is the moment a host has just launched us from an npm package or a Homebrew cellar, and "the docs were next to the binary" is not a thing that survives packaging. A prompt that is sometimes missing is worse than no prompt: the agent's behaviour then depends on how the user installed us.

Three ways out, one source:

• the MCP `initialize` response's `instructions` field, which every conforming host puts in front of its model without the user doing anything at all. This is the one that matters; the others are for people who want the text somewhere they can edit it.
• `brain prompt`, to paste into a CLAUDE.md or a system prompt.
• systemmd/BRAINPROMPT.md in the repository, which is the same bytes.

Index

• func Text() string

Functions

func Text

func Text() string

Text is the full instruction document.