Engineering Docs

22 composable, auto-triggering skills that turn your coding agent into a principal engineer - from raw idea to production-ready documentation.

Why Engineering Docs?

Your coding agent is powerful, but it doesn't know your project's architecture, your users, or your constraints. Engineering Docs gives it principal-level documentation skills — so it can:

• Turn raw ideas into complete blueprints — business concept → technical spec → architecture → deployment plan
• Ask the right questions — tool-call interviews with 2-3 targeted questions per skill (no repeated questions)
• Generate production-ready documents — ISO/IEC/IEEE 29148, C4 Model, STRIDE, Google SRE standards
• Work across 14+ agents — Claude Code, Copilot, Cursor, Gemini CLI, Goose, Pi, and more

How It Works

The plugin works through a structured workflow:

1. User gives idea → Orchestrator skill activates automatically
2. Mode detection → Greenfield (new) vs Brownfield (existing)
3. Interview phase → Tool-call questions with context loading
4. Document generation → Sequential generation with 22 specialized skills
5. Consistency checks → Cross-document verification
6. Master index → Complete blueprint ready for implementation

Quickstart

npx engineering-docs

Or install for your specific agent:

See Installation for detailed instructions.

How It Works

graph LR
    A[Your Idea] --> B[Orchestrator]
    B --> C[Interview]
    C --> D[Business Concept]
    D --> E[Project Plan]
    E --> F[Technical Spec]
    F --> G[System Architecture]
    G --> H[API Design]
    H --> I[Implementation Plan]
    I --> J[Test Strategy]
    J --> K[Deployment Plan]
    K --> L[Master Index]

1. Give it your idea — "I want to build X"
2. Answer 2-3 questions per skill — via tool calls, not inline chat
3. Review each document — approve or request changes
4. Get your blueprint — complete, cross-consistent documentation set

Smart features:

• Context loading — reads prior documents before asking questions (never repeats)
• Tool-call interviews — clean input capture, no conversation pollution
• Right-sizing — skips documents that don't apply to your project
• Cross-document consistency — verifies entity names, roles, decisions match

What's Inside

Skills Library (22 Skills)

Discovery & Planning

Specification & Feasibility

Architecture & Design

Quality & Risk

Delivery & Operations

Custom Agents (4 Agents)

Utility Scripts

MCP Integration

.mcp.json          # MCP server configuration
scripts/validate.js    # Validation server (validate_document_set, check_consistency, generate_index)

Hooks

hooks/hooks.json          # SessionStart hook configuration
hooks/check-progress.js   # Check for in-progress documentation
hooks/run-hook.cmd        # Windows compatibility

Eval Framework

evals/evals.json           # Test cases for orchestrator and key skills
evals/README.md            # How to run evals
evals/test-prompts/        # Sample test prompts

Plugin Structure (Claude Compatible)

engineering-docs/
├── .claude-plugin/
│   ├── plugin.json        # Plugin manifest
│   └── marketplace.json   # Marketplace manifest
├── skills/                # 22 skills (SKILL.md files)
├── agents/                # 4 custom agents
├── hooks/                 # Event handlers
├── .mcp.json              # MCP server configuration
├── scripts/               # All scripts (install, validate, test, utilities) (install.js, validate.js, test-skills.js)
├── scripts/               # Utility scripts + setup scripts
├── evals/                 # Test framework
└── integrations/          # Other agent platform configs
    ├── agents/            # Agent configs (AGENTS.md, CLAUDE.md, GEMINI.md)
    └── plugins/           # Plugin configs for 13+ platforms

Claude Guidelines Compliance:

• ✅ Components at plugin root (not inside .claude-plugin/)
• ✅ Skills in skills/ directory with SKILL.md
• ✅ Agents in agents/ directory with frontmatter
• ✅ Hooks in hooks/hooks.json
• ✅ MCP in .mcp.json
• ✅ Kebab-case naming
• ✅ Validation passes: claude plugin validate .

Installation

CLI Installer (Recommended)

npx engineering-docs

Automatically detects and copies the plugin to your agent's directory.

Per-Agent Installation

Claude Code

# Official marketplace
/plugin install engineering-docs@claude-plugins-official

# Or register marketplace first
/plugin marketplace add fattain-naime/engineering-docs
/plugin install engineering-docs@engineering-docs

Gemini CLI

gemini extensions install https://github.com/fattain-naime/engineering-docs

Or clone manually:

git clone https://github.com/fattain-naime/engineering-docs.git ~/.gemini/config/plugins/engineering-docs

Cursor / Windsurf

npx engineering-docs --cursor

Extracts each skill to .cursor/rules/engineering-docs-*.mdc.

Goose

npx engineering-docs --goose

Or configure manually in ~/.config/goose/config.yaml.

Pi

npx engineering-docs --pi

Or install from git:

pi install git:github.com/fattain-naime/engineering-docs

OpenCode

npx engineering-docs --opencode

Kilo Code

npx engineering-docs --kilo

Or install from Kilo Code plugin marketplace.

Codex / GitHub Copilot

npx engineering-docs --codex

Copilot CLI

npx engineering-docs --copilot

Cline

npx engineering-docs --cline

Copies .clinerules to your project root.

Factory Droid

npx engineering-docs --factory

Roo Code

npx engineering-docs --roo

Kimi Code

npx engineering-docs --kimi

Or install inside Kimi Code:

/plugins install https://github.com/fattain-naime/engineering-docs

Cross-Platform Scripts

# Windows (PowerShell)
pwsh scripts\setup.ps1
pwsh scripts\setup.ps1 -Target gemini
pwsh scripts\setup.ps1 -Target claude

# Linux / macOS (Bash)
chmod +x scripts/setup.sh
./scripts/setup.sh
./scripts/setup.sh --gemini
./scripts/setup.sh --claude

Supported targets: gemini, claude, local, cursor, kimi, codex, goose, pi, opencode, kilo, roo, cline, factory, copilot

Safe-Write Behavior

All install methods use safe-write for agent config files:

• AGENTS.md — Created only if it doesn't exist
• CLAUDE.md — Created only if it doesn't exist
• GEMINI.md — Created only if it doesn't exist
• COPILOT.md — Created only if it doesn't exist
• GOOSE.md — Created only if it doesn't exist
• PI.md — Created only if it doesn't exist

Your customizations are always preserved.

Multi-Agent Compatibility

Testing

# Run plugin validation
claude plugin validate .

# Run skill tests
npm test

# Test MCP server
node scripts/validate.js

Philosophy

• Systemic over Ad-hoc — Rigorous, reproducible processes yield safer and cleaner software
• Traceability — Every requirement links to a business goal and a test case
• Visual-First — Complex architectures mapped with Git-trackable Mermaid diagrams
• Operational Safety — No feature is complete without deployment runsheet, monitoring, rollback
• Blameless Learning — Production failures are data points for system hardening

Contributing

We welcome community skills! Please review CONTRIBUTING.md for guidelines.

Community

• GitHub: fattain-naime/engineering-docs
• Issues: Report bugs or request features

Author

• Fattain Naime — iamnaime.info.bd
• Repository: https://github.com/fattain-naime/engineering-docs

License

MIT. See LICENSE.