sharptoolbox/ontology-driven-dev

A set of ontology-driven business system development skills using a three-step approach of “requirements exploration → ontology modeling → application construction.” Based on the seven model ontologies YAML (M1/M2/M3/M5/M6/M7/MU) and the built-in [T0]] technology base, manual confirmation is required at each requirement exploration stage, ultimately producing a runnable browser front-end and back-end (BS) system that strictly aligns requirements documents, ontology models, and code**.

English version → README_EN.md

---

1. What can this skill do?

Turn a business requirement (a sentence or a description) into a runnable management system, and guarantee:

• Requirement Traceability: Each feature can be traced back to a specific entry in the requirements document;
• Model as Semantic Source: Database tables, interfaces, menus, permissions, processes, and rules are all generated by seven models using YAML, eliminating the principle of “one model, one code”;
• Manual access control cannot be skipped: The 8 stages of requirement exploration, each must be paused and confirmed by the person before progressing.
• Out-of-the-box tech foundation: Built-in code-paas (Flask + SQLite + React/TS monolithic application, including system management, process engine, workbench, ontbook registry), can be copied and expanded;
• Forced AI Dialog Box: The generated application comes with AI dialogue on the right side (ontology registry injection + tool call + SSE streaming + read-only SQL security boundary).

---

2. Include content

Complete application example

code-app-example/ Presented the complete implementation results from requirement specifications and the seven-model ontology to the runnable system. The sample business is sales contract execution management, covering contract registration and hierarchical approval, invoicing approval, collection and write-off, cross-object status linkage, fixed reports, system management, and the AI intelligent assistant on the right.

The sample includes SQLite test data and default test accounts, allowing you to directly start and verify the complete business chain according to the instructions in code-app-example/README.md. The AI API Key in the warehouse has been cleared. Before using the intelligent assistant, you need to configure OpenAI-compatible large model services in the system interface.

---

3. Installation (Mainstream Tools)

This skill does not rely on any WorkBuddy proprietary mechanisms and can fully run on tools such as Claude Code, Codex, Cursor, etc. Unique adaptation point: Within skills, the relative path is rooted in the folder where the SKILL.md is located, and each tool will automatically parse it.

1. WorkBuddy

You can simply say the trigger in the WorkBuddy conversation (see Section 5).

2. Claude Code

The Skills format of Claude Code is exactly the same as the frontmatter (name / description) of this skill, so basically, just copy directly:

Claude Code automatically discovers .claude/skills/ /SKILL.md and executes it according to its flow.

3. Codex