AI Development Introduction

This page is the entrypoint for contributors who are designing, reviewing, or maintaining AI vibe coding workflows in the Cabloy repository.

The Cabloy monorepo is a good fit for AI vibe coding because the source tree already contains most of the framework knowledge an agent needs:

• root scripts for shared workflows
• Vona CLI source and command groups
• Zova CLI source and command groups
• archived docs that still capture valuable concepts
• internal engineering docs for maintainers
• Claude commands and skills

The main design goal

The goal is not to make AI guess Cabloy conventions more effectively.

The goal is to make AI reuse the repo’s existing conventions directly, especially through:

• CLI commands
• root scripts
• repo markers
• internal architecture notes
• shared public documentation

The knowledge layers

Public docs

Use cabloy-docs/ for user-facing and agent-facing guidance that should remain durable and source-aligned.

For normal project usage, prefer the user-facing entry docs such as Fullstack Quickstart. This AI section focuses on repository workflows and AI vibe coding.

Internal engineering docs

Use .docs-internal/ for architecture notes, ADRs, and maintainership rationale that should not be mixed into public how-to documentation.

Claude rules and commands

Use root CLAUDE.md and .claude/commands/ for concise operational behavior and repeatable workflows.

Skills

Use .claude/skills/ for procedural workflows that benefit from reusable instructions, bundled references, or future deterministic scripts.

Recommended AI lookup rules

For backend shorthand lookup, prefer these surfaces in order:

1. IBeanRecordGlobal for this.bean.xxx, ctx.bean.xxx, and app.bean.xxx
2. module src/.metadata/index.ts to map shorthand names to generated bean types
3. src/bean for the shorthand source file itself
4. only then re-evaluate whether the target is really a general full-name bean, a service-scene runtime-anchor, or a lib/helper class

For the full lookup workflow, read Global Bean Lookup.

Common AI mistakes to avoid

• assuming Cabloy Basic and Cabloy Start are identical
• creating framework files manually when a CLI command already exists
• trusting stale legacy repo paths instead of current source
• mixing public documentation and internal rationale into one document set