Skip to main content

​
Installation

Get the SDD plugin up and running in your Claude Code environment.

​
Prerequisites

Before installing the SDD plugin, ensure you have:
1

Claude Code installed

You need Claude Code installed and running
2

Node.js 18+ (optional)

Required for the MCP server and context augmentation featuresCheck your version:
node --version
3

Git (recommended)

Required for commit traceability featuresVerify installation:
git --version
The plugin works without Node.js and Git, but some features will be unavailable.

​
Install the plugin

Inside Claude Code, run the installation command: 
/install-plugin github:noelserdna/claude-plugin-sdd
Claude Code will download and activate the plugin automatically. You should see a confirmation message once installation completes.

​
Verify installation

Check that the plugin is active: 
/plugin list
You should see sdd in the list of installed plugins.
You can also verify by checking if SDD skills are available. Try typing /sdd: and you should see autocomplete suggestions for all 20 skills.

​
Initialize your project

Navigate to your project directory and run the setup skill: 
/sdd:setup
This command will:
  1. Verify the plugin is active
  2. Create pipeline-state.json in your project root
  3. Check for dependencies (jq, node)
  4. Build the MCP server (if Node.js is available)
  5. Report the status of all components

​
Expected output

After running setup, you’ll see a status table like this: 
## SDD Pipeline Initialized

| Component | Status |
|-----------|--------|
| Plugin: sdd | Active |
| Pipeline: pipeline-state.json | Initialized |
| MCP Server: server/dist/ | Built |
| Dependency: jq | Available |
| Dependency: node | Available |
If jq is missing, hooks will fall back to node -e (slower but functional). If Node.js is missing, the MCP server won’t be available.

​
What gets installed

The SDD plugin includes several automatic components:

​
Hooks (run every session)

  • H1 — Injects pipeline status at session start
  • H2 — Blocks downstream skills from editing upstream artifacts
  • H3 — Auto-updates pipeline state when artifacts change
  • H4 — Consistency check at session end
  • H5 — Context augmentation (injects SDD traceability into file reads)

​
Agents (delegated by Claude)

  • A1-A3 — Constitution enforcer, cross-auditor, context keeper
  • A4-A8 — Requirements watcher, spec compliance, test coverage, traceability validator, health monitor

​
MCP Server

5 tools for live traceability queries:
ToolPurpose
sdd_querySearch artifacts by text, ID, type, or domain
sdd_impactBlast radius analysis by depth
sdd_context360° view of any artifact
sdd_coverageGap analysis by domain or layer
sdd_traceFull chain traversal with break detection
The MCP server reads dashboard/traceability-graph.json generated by /sdd:dashboard. Run the dashboard skill first to populate data.

​
Project structure

After initialization, your project will have this structure: 
your-project/
├── pipeline-state.json          # Pipeline progress tracking
├── requirements/                # (created by requirements-engineer)
├── spec/                        # (created by specifications-engineer)
├── audits/                      # (created by spec-auditor)
├── test/                        # (created by test-planner)
├── plan/                        # (created by plan-architect)
├── task/                        # (created by task-generator)
├── src/                         # (created by task-implementer)
├── tests/                       # (created by task-implementer)
└── dashboard/                   # (created by dashboard skill)
    ├── index.html
    ├── guide.html
    └── traceability-graph.json
Directories are created automatically as you progress through the pipeline. You don’t need to create them manually.

​
Optional features

​
Code intelligence

For deep traceability with symbol-level analysis: 
/sdd:code-index
Run this after /sdd:dashboard to enable code-aware blast radius analysis. Works standalone but produces richer results with GitNexus installed.

​
Notion integration

For bidirectional sync with Notion databases: 
/sdd:sync-notion
You’ll need to provide your Notion API token and database IDs.

​
Troubleshooting

​
Plugin not found

If /plugin list doesn’t show sdd:
  1. Restart Claude Code
  2. Try reinstalling: /install-plugin github:noelserdna/claude-plugin-sdd
  3. Check your internet connection (plugin downloads from GitHub)

​
MCP server failed to build

If the MCP server doesn’t build:
  1. Verify Node.js 18+ is installed: node --version
  2. Navigate to the plugin directory and manually run: 
    cd ~/.config/claude-code/plugins/sdd/server
npm install
npm run build

​
pipeline-state.json not created

If setup doesn’t create the state file:
  1. Verify you’re in a project directory (not the plugin directory)
  2. Check you have write permissions in the current directory
  3. Manually create the file by running /sdd:setup again

​
Next steps

Quick start

Run your first pipeline from requirements to code

Pipeline overview

Learn about the 7 sequential pipeline steps

Build docs developers (and LLMs) love

Get started for freeTalk to us