sdd-kit 2.2.6

Spec-Driven Development CLI with IDE-native AI

sdd-kit v2.0.6

The Spec-Driven Development (SDD) Toolkit for AI-Native Engineering. Intelligently assemble AI context from enterprise data lakehouses (Databricks/Snowflake) directly into your IDE.

---

🚀 Quick Install

# 1. Install the CLI
pip install sdd-kit

# 2. Install the VS Code extension (No API key required!)
sdd install-extension

---

✨ What makes sdd-kit different?

sdd-kit isn't just another AI tool. It's a context orchestration engine that bridges the gap between your enterprise data (Gold/Silver schemas) and your AI agent (Cursor/Copilot/ChatGPT).

• IDE-Native AI: Use your existing IDE's AI credentials. No more ANTHROPIC_API_KEY errors in the terminal.
• Lakehouse Aware: Pull live schema definitions and documentation from Databricks or Snowflake.
• Token Budgeting: Automatically fits massive codebases into tight 8,000-token context windows.
• Zero-Extension Mode: Full support for Cursor/Windsurf via MCP (Model Context Protocol).

---

🛠️ Main Workflows

1. Existing Projects (Audit & Evolve)

Perfect for onboarding a "maintaining" project into SDD without touching existing files.

cd my-legacy-project
sdd onboard          # surgically adds .sdd/ and .cursor/
@sdd /audit          # Run in VS Code Chat to audit the codebase
@sdd /specify-next   # Define a new feature based on existing code

2. New Projects (Spec-to-Code)

Initialize a project from a domain-specific "Gold" knowledge base.

sdd init my-app --domain banking
@sdd /specify        # Generate a perfect spec.md
@sdd /plan           # Generate plan.md and tasks.md

---

💬 Slash Commands

When you use VS Code (after running sdd install-extension) or Cursor, you get these powerful commands directly in your AI Chat panel:

Command	Purpose
/audit	Scan existing code for architecture and technical debt
/specify	Generate a comprehensive master specification
/plan	Create a step-by-step implementation plan and checklist
/integrate	Reverse-engineer or generate live Lakehouse integrations
/doctor	Run diagnostics on your local development environment
/sync-kb	Mirror enterprise knowledge for offline development

---

🛡️ The 12 SDD Rules

The toolkit enforces a rigorous methodology for AI-assisted engineering:

1. Think Before Coding — State assumptions. Ask if unclear.
2. Simplicity First — Minimum code that solves the problem.
3. Surgical Changes — Touch only what the task requires.
4. Context Budget Discipline — No full files. 8K max tokens.
5. Existing Projects Are Not Broken — Recommend delta only.
6. Offline is First-Class — All commands work without internet.

---

🔧 Configuration (Optional)

If you want to pull live data from your lakehouse, set these variables:

# Databricks
export DATABRICKS_HOST=https://adb-xxx.azuredatabricks.net
export DATABRICKS_TOKEN=dapi...

# Snowflake
export SNOWFLAKE_ACCOUNT=xxx
export SNOWFLAKE_USER=xxx

---

📄 License

MIT © 2026 Your Company