01
Documentation before drift
Architecture and behavior are documented as contracts instead of being reconstructed after implementation changes.
Documentation-first AI engineering toolkit
Build AI-assisted development workflows that are explicit, reusable, and maintainable.
Qbit AI Toolkit is the versioned home for prompt-engineering guidance, AI development tooling, MCP usage, reusable agent skills, policies, templates, and the engineering contracts that connect them.
One toolkit, clear boundaries
Learn the method, use the tools, understand the contracts.
Each documentation domain has its own navigation so operational tooling, prompt design, MCP integrations, agent behavior, and engineering reference do not collapse into one mixed hierarchy.
PE
Prompt Engineering
Learn how to write explicit prompts, use reusable patterns, and evaluate behavior instead of relying on ad-hoc wording.
Explore section
AI
AI Tools
Setup scripts, reusable assets, Codex AI Tooling, repository-local helpers, verification, maintenance, and troubleshooting.
Explore section
MCP
MCP
Configure and use Model Context Protocol integrations with explicit capability, permission, transport, and security boundaries.
Explore section
AS
Agents & Skills
Design reusable skills, publish ready-made workflows, and separate agent policies from task-specific execution logic.
Explore section
ENG
Engineering Reference
Architecture, asset contracts, conventions, security boundaries, versioning, and compatibility rules for maintainers.
Explore section
Current operational asset
Explore Codex AI Tooling Codex AI Tooling
The first implemented catalog asset installs repository-owned AI development tooling through a versioned, cross-platform installer lifecycle with explicit ownership, verification, doctor, recovery, and uninstall behavior.
Project principles
Designed for production-grade engineering workflows.
02
Reusable without hidden coupling
Prompts, skills, templates, installers, and policies have explicit ownership and consumer boundaries.
03
Validation is part of the asset
Operational tooling is expected to expose verification, failure behavior, and evidence—not just create files.
Start from the area that matches the problem you are solving.
Use the top navigation to move between learning material, operational tooling, integrations, reusable agent behavior, and engineering reference.