Init Project

init-project

Initialize any software project or repository according to the Production Multi-Agent Governance Standard.

Sets up universal agent governance (AGENTS.md, GEMINI.md, CLAUDE.md), dynamic architecture specifications (ARCHITECTURE.md), context scratchpads (.vault-notes.md), lightweight pre-commit security gates, zero-pollution analytics isolation, and browser runtime verification protocols.

Trigger

Run whenever the user or agent initiates project setup:

• init project, init-project, initialize project, setup project
• configure project rules, bootstrap agent governance, setup agents
• When opening a new repo or workspace and establishing AI agent governance rules.

What this skill does

1. Scans target workspace: detects language and framework (Node/Astro/Next/React/Python/FastAPI/Docker), existing files, and Git status.
2. Classifies project profile:
   • Public Web / SaaS / Client Site: full SEO, security headers, cookie/consent gating, and contact intake validation.
   • Internal App / CLI Tool / Microservice: streamlined auth, focuses on API contracts and performance, skips consumer cookie banners.
3. Generates Universal Multi-Agent Instructions:
   • AGENTS.md (Master operational blueprint for Claude Code, Antigravity, Gemini, Cursor, and Windsurf).
   • GEMINI.md (Reference linking to AGENTS.md).
   • CLAUDE.md (Reference linking to AGENTS.md).
4. Generates Spec-Driven Documentation:
   • ARCHITECTURE.md (Living technical spec: directory tree, component contracts, schema, endpoints).
   • README.md (Quickstart, verification commands, and high-level architecture overview).
   • Enforces the Spec-Driven Rule: code changes must update ARCHITECTURE.md.
5. Configures Context Scratchpad & Memory:
   • Initializes .vault-notes.md for intra-session continuity.
   • Updates .gitignore to prevent committing local scratchpads and secrets.
   • (Optional) Connects Second Brain / Obsidian knowledge base synchronization.
6. Injects Lightweight Quality & Security Gates:
   • Automated syntax & build validation commands.
   • Fast pre-deploy secret leak check (detecting .env leaks, private tokens, honeypots).
   • Analytics zero-pollution guardrail (preventing local/test traffic from contaminating production analytics).
   • Browser runtime verification protocol via Chrome DevTools or Playwright.

Execution Steps

Step 1 — Project Discovery & Profile

Inspect the workspace:

• Check root files:
  • package.json -> inspect dependencies (astro, next, react, vue, express, fastify).
  • requirements.txt / pyproject.toml -> Python environment.
  • Dockerfile / docker-compose.yml -> Containerized service.
  • Verify if Git is initialized (git rev-parse --is-inside-work-tree).

Determine variables:

• PROJECT_NAME: directory basename.
• PROJECT_TYPE:
  • Public Web Platform / Client Site
  • Internal Tool / Microservice / API
• BUILD_COMMAND: e.g. npm run build, npm test, or pytest.

Step 2 — Generate AGENTS.md

Create AGENTS.md in the project root:

# AI Agent Instructions — PROJECT_NAME

> Project name: `PROJECT_NAME`
> Target environment: [Production URL / Staging URL / Localhost]

---

## 🎯 Project Overview & Core Focus

[Brief 2-3 sentence summary of what this project is, its core capabilities, and tech stack.]

---

## 🚨 Mandatory Project Documentation Rule (Spec-Driven)

**Whenever making ANY changes to the codebase, API routes, data models, or client forms:**
1. **Always update [`ARCHITECTURE.md`](ARCHITECTURE.md)** detailing modified/new components, endpoints, schema changes, security mechanisms, and data flows.
2. **Always update [`README.md`](README.md)** with user-facing summaries, API tables, and operational guides.
3. Keep documentation strictly synchronized with reality. Never leave stale architecture descriptions.

---

## 🧪 Verification & Runtime Testing Protocol (Mandatory)

**Never claim work is complete without rigorous verification.**

Before finishing any task or asking for review:
1. **Syntax & Build Validation:**
   - [SPECIFIC_BUILD_COMMAND, e.g. `npm run build` or `python3 -m py_compile ...`] (must pass with 0 errors).
2. **Automated Test Suite:**
   - [SPECIFIC_TEST_COMMAND, e.g. `npm test` or `pytest`] (must pass 100% of all tests).
3. **Browser Runtime Verification (Chrome DevTools / Playwright):**
   - Open the local dev/preview URL.
   - Check console messages: **0 unhandled JavaScript errors** (`TypeError`, `ReferenceError`).
   - Verify layout responsiveness (mobile < 480px and desktop > 1200px).
   - Verify dynamic element rendering.
4. **Analytics Zero-Pollution Guardrail:**
   - All tests, agents, and local development MUST suppress analytics tracking.
   - Ensure `localhost`, `127.0.0.1`, `navigator.webdriver === true`, `?admin=1`, or `localStorage.getItem('admin_ga_disabled')` prevents sending pageviews/events to GA4/Clarity.

---

## 🔒 Pre-Deploy Security Checklist (Fast & Low-Cost)

Before committing or pushing to production:
1. **Zero Leaked Secrets:** Check `git status` to ensure `.env` and local secrets are NOT tracked. Scan for hardcoded keys (`sk-`, `AIza`, `ghp_`, bearer tokens).
2. **Form Sanitization & Honeypots:** Ensure all customer-facing forms have active bot honeypots and input sanitization.
3. **Clean Style & Content:** In copywriting and docs, do NOT use em dashes `—` or AI buzzwords ("game-changer", "dive in", "delve", "tapestry").

> **Note on Major Launches:** For full production milestones or public releases, run the comprehensive audit skill: `prod-readiness-audit`.

---

## 🔄 Session Protocol (Context Continuity)

Follow this 3-phase protocol on every work session:

### Phase 1 — Session Start
1. Check if `.vault-notes.md` exists in this folder.
   - If yes: read it to absorb previous context.
   - If no: create it with today's date and a brief note about the current task.

### Phase 2 — During Session
Keep `.vault-notes.md` updated after each meaningful change:
- What was built or changed, in plain language.
- The decision or reason behind it.
- What is incomplete or next.

### Phase 3 — Before Commit / End of Session
Do this BEFORE committing:
1. Summarize key achievements and write a clean, structured commit message.
2. Clear `.vault-notes.md` (or let the post-commit hook archive it automatically).

Step 3 — Generate GEMINI.md and CLAUDE.md

Create GEMINI.md:

# Gemini & Antigravity Instructions

See [AGENTS.md](AGENTS.md) — all core instructions, testing rules, and session protocols apply here.

Mandatory rules:
1. Always maintain and update `ARCHITECTURE.md` and `README.md` with all technical specifications.
2. Run syntax/build and test commands before claiming work is complete.
3. Verify visual UI runtime with 0 unhandled console errors.

Create CLAUDE.md:

# Claude Code Instructions

See [AGENTS.md](AGENTS.md) — master instructions apply to all sessions.

Follow the 3-phase session protocol (`.vault-notes.md` context continuity).
Always update `ARCHITECTURE.md` and verify 0 console errors before completing tasks.

Step 4 — Generate ARCHITECTURE.md Blueprint

Create ARCHITECTURE.md:

# Architecture Specification: PROJECT_NAME

> **Status:** Active Development  
> **Environment:** [URL / Localhost]  
> **Repository:** [Git repository URL]  

---

## 1. Concept & Scope

High-level summary of the system, business problems solved, and primary personas.

---

## 2. Directory Structure

PROJECT_NAME/ ├── ARCHITECTURE.md # Complete technical blueprint ├── README.md # User guide & quickstart ├── AGENTS.md # Universal multi-agent governance ├── GEMINI.md # Reference to AGENTS.md ├── CLAUDE.md # Reference to AGENTS.md ├── .vault-notes.md # Session context scratchpad │ ├── [src/ / app/ / core/] # Application source code └── tests/ # Automated test suite

---

## 3. Core Components & API Endpoints

| Route / Component | Purpose | Access |
|---|---|---|
| `/` | Main UI / Entry point | Public |
| `/api/health` | Service health check | Public |

---

## 4. Data Models & State Management

Database schema, migrations, tables, or client state handling.

---

## 5. Integrations & External Services

- **Webhooks & APIs:** Downstream event pipelines.
- **Observability:** Logging, analytics, and error tracking.

---

## 6. Security & Environment Configuration

- Secrets managed via `.env` (never committed).
- Bot honeypots on intake forms.
- Rate-limiting and input sanitization.

Step 5 — Generate README.md

Create or update README.md:

# PROJECT_NAME

[One-line elevator pitch for the project]

## Quickstart

```bash
# Install dependencies
npm install  # or pip install -r requirements.txt

# Run development server
npm run dev

# Run verification and build
npm test
npm run build

Documentation & Architecture

• Technical specification and architecture: ARCHITECTURE.md
• Multi-agent instructions and testing protocols: AGENTS.md

Development Rules

1. Any changes in code, schema, or routes must be reflected in ARCHITECTURE.md.
2. Ensure 0 unhandled console errors and clean test runs before committing.
3. Pre-deploy security checks are mandatory before production releases.

---

### Step 6 — Context Scratchpad & Git Configuration

1. **Create `.vault-notes.md`:**
```markdown
# Context Notes — PROJECT_NAME

Initialized: YYYY-MM-DD

[Active session context will be maintained here.]

2. Update .gitignore: Ensure the following entries are ignored:

.vault-notes.md
.vault-notes-archive/
.env
.env.*
!.env.example

3. (Optional) Install Post-Commit Archival Hook: To automatically archive .vault-notes.md on every commit, copy the provided hook:

mkdir -p .git/hooks
cp templates/post-commit.sh .git/hooks/post-commit
chmod +x .git/hooks/post-commit