This guide explains how to properly configure AI coding assistants to work with the BTC-Koii repository in a way that adheres to project directives and plans.
Claude Code is Anthropic's command-line AI assistant that can help with coding tasks while adhering to project guidelines.
-
Initialization:
When starting a new Claude Code session, provide the configuration file:
claude-code --config-file claude-code.config.json
-
Guide Claude:
At the beginning of your session, inform Claude Code about the project directives:
Before helping me with this project, please read the directives in /directives/*.md and ensure all suggestions follow the project plan and directives. -
Reference Mode:
Enable reference mode to ensure Claude Code references plan documents:
Please use reference mode and cite specific sections of the plan or directives when explaining your implementations.
Cursor is an AI-powered code editor that can be configured to follow project directives.
-
Import Configuration:
In Cursor, navigate to Settings > AI Configuration and load the
cursor.config.jsonfile. -
Project Initialization:
When opening the project in Cursor, run this command in the AI chat:
/analyze-project /read-files directives/*.md plan.md phase-0/plan.md -
Working Mode:
Set the AI to strict mode:
/set-constraints "Follow BTC-Koii directives. Create plan_update.md for deviations. Never modify master plan."
The repository includes KNO SDK integration that can be used to give AI agents better context about the project:
-
Generate Embeddings (if not already present):
python generate_kno_embeddings.py
-
Reference in AI Tools:
Both Claude Code and Cursor configurations are set to use these embeddings for context.
When working with AI assistants on this project:
-
Always mention the plan:
I want to implement feature X, which is specified in phase-0/plan.md section Y -
For deviations, explicitly request documentation:
We need to change approach for feature X due to obstacle Z. Please help me document this in a plan_update.md file in the phase directory. -
Validate against directives:
Please ensure this implementation follows all directives, especially prevent_hallucinations.md -
Request plan references:
Explain how this implementation fulfills the requirements in the plan
Following these guidelines ensures AI assistants remain aligned with project directives and produce code that adheres to the established plan structure.