Skip to main content

Project Setup + Constitution

You're building a Personal AI Business Intelligence System with four features: Brand Profiler, Market Scanner, Content Strategy, and Action Dashboard. Before using any AI tools, you'll initialize your project with Spec-Kit Plus and create a constitution using /sp.constitution.

This constitution becomes your decision framework—the standard you reference when building each feature. If Feature 4 builds significantly faster than Feature 1, you'll know it's because you stopped re-deciding quality standards and reused your constitution.

Gather Your Personal Data​

Before anything else, collect the raw material for your brand analysis.

Open these in browser tabs:

  1. LinkedIn — Your profile's "About" section
  2. GitHub — Your profile bio and pinned repositories
  3. Portfolio (if you have one) — Your main description/tagline

Copy this data somewhere accessible — You'll paste it into prompts throughout this capstone.

Minimum required:

  • LinkedIn About section (1-3 paragraphs)
  • GitHub bio (1-2 sentences)
  • Target role/industry (what kind of roles are you targeting?)
  • 3-5 target companies (companies you'd want to work for)

Initialize Your Project​

Use Spec-Kit Plus to create the project structure. In your terminal:

# Create a new Spec-Kit Plus project
specifyplus init personal-bi-system

# Navigate to the project
cd personal-bi-system

Interactive Prompts:

During initialization, you'll see:

? Select AI Tool:
  > Claude Code
    Gemini CLI

? Select Terminal:
  > bash
    powershell (Windows only)

Recommendations:

  • AI Tool: Choose Claude Code (recommended for this book)
  • Terminal: Choose bash (or powershell if on Windows without WSL)

Verify your project structure:

ls -la

Expected:

personal-bi-system/
├── .claude/
│   └── commands/           # Slash commands (/sp.specify, /sp.plan, etc.)
├── .specify/
│   ├── memory/             # Constitution lives here
│   ├── scripts/            # Automation scripts
│   └── templates/          # Templates for specs, plans, tasks
├── CLAUDE.md               # Agent instructions
├── README.md               # Project documentation
└── .gitignore

The specs/ and history/ directories will be created automatically when you start your first feature.

Create Your Constitution​

The constitution defines quality standards for all four features. Use /sp.constitution to create it interactively.

In Claude Code, run:

/sp.constitution

Project principles and standards:

Vision: Build an intelligent system that analyzes your professional presence, researches your target market, and generates actionable career positioning strategy—all using AI tools (Gemini App, NotebookLM) without writing code.

Core Principles:
1. Your Real Data: All analysis uses YOUR actual profiles and target companies—not hypothetical data.
2. Structured Outputs: Every feature produces structured markdown with clear sections. No freeform text.
3. Actionable Results: Every output includes specific next steps. Analysis without action items is incomplete.
4. Source Traceability: Market research cites specific sources. Brand analysis references specific profile sections.

Technical Stack:
- Gemini App (gemini.google.com) for brand analysis and content generation
- NotebookLM (notebooklm.google.com) for multi-source research synthesis
- Claude Code for SDD-RI workflow orchestration
- No student-written code (AI tools do the work)

Feature Specifications:

### Feature 1: Personal Brand Profiler
Tool: Gemini App
Input: LinkedIn About, GitHub bio, portfolio description
Output: Brand analysis with 3+ strengths (with evidence citations), 2+ gaps, positioning statement, differentiation opportunities, confidence score
Quality Gate: Must cite specific phrases from profile as evidence for each strength

### Feature 2: Market Intelligence Scanner
Tool: NotebookLM
Input: 3-5 target company pages, job postings, industry articles
Output: Market brief with 3+ trends, in-demand skills, competitor landscape, opportunity areas, source citations
Quality Gate: Must synthesize 3+ different sources with specific citations

### Feature 3: Content Strategy Generator
Tool: Gemini App
Input: F1 brand analysis + F2 market brief
Output: Content strategy with 3 pillars, weekly schedule, 10+ topics, format recommendations, first week actions
Quality Gate: Content pillars must connect F1 strengths to F2 market trends

### Feature 4: Action Dashboard
Tool: Claude Code (markdown aggregation)
Input: F1 + F2 + F3 outputs
Output: Unified dashboard with brand summary, market opportunities, content calendar, top 5 actions, 30/60/90 day goals
Quality Gate: Dashboard aggregates all three features without redundancy. Each section traces to source feature.

Non-Goals:
- Production web application
- Automated social media posting
- Database or persistent storage beyond files
- Real-time market data feeds

Constraints:
- Tools: Only Gemini App, NotebookLM, Claude Code
- Data: Only public information + your own profiles
- No code: Students don't write Python (that's Part 5)
- Time: Each feature should be < 60 minutes (F4 target: < 30 minutes)

Agent does:

  • Creates .specify/memory/constitution.md with your project standards
  • Sets up quality gates for each feature
  • Defines tool constraints and non-goals

Verify the constitution:

cat .specify/memory/constitution.md

Read through it once. This is your quality gate document for all four features.

Set Up Time Tracking​

Create a time tracker to measure intelligence accumulation:

Ask Claude Code:

Create a TIME_TRACKER.md file in the project root with this structure:

# Feature Build Times - Personal BI System

| Feature | Start Time | End Time | Duration (min) | Notes |
|---------|------------|----------|----------------|-------|
| F1: Personal Brand Profiler | | | | |
| F2: Market Intelligence Scanner | | | | |
| F3: Content Strategy Generator | | | | |
| F4: Action Dashboard | | | | |

## Acceleration Analysis

After all four features:
- F1 baseline: _____ minutes
- F4 actual: _____ minutes
- **F4 vs F1 ratio**: _____ % (calculate: F4_time / F1_time × 100)
- **Target: F4 ≤ 50% of F1**

## Pattern Transfer Notes

What patterns from earlier features made later features faster?
- Prompting patterns reused:
- Output validation shortcuts:
- Quality gate checks that became automatic:

Verify Complete Setup​

Check your project structure:

ls -la
ls .specify/memory/

Expected files:

  • .specify/memory/constitution.md — Quality standards (created by /sp.constitution)
  • TIME_TRACKER.md — Acceleration measurement
  • .claude/commands/ — Slash commands for SDD workflow

Commit Your Constitution​

Before starting features, commit the constitution to git:

/sp.git.commit_pr Commit the constitution as the foundation for the Personal BI project.

Agent does:

  • Commits constitution and project structure
  • Creates PR or pushes to branch

What's Next​

You now have:

  1. Spec-Kit Plus project initialized with correct structure
  2. Constitution defining quality gates for all four features
  3. Time tracker ready to measure acceleration
  4. Personal data gathered for Feature 1

In Lesson 2, you'll build Feature 1 (Personal Brand Profiler) using the full SDD-RI workflow:

  • /sp.specify — Create specification
  • /sp.plan — Generate implementation plan
  • /sp.tasks — Break into atomic tasks
  • /sp.implement — Execute with Gemini App

Start your F1 timer when you begin Lesson 2.