razbakov/portfolio-hiring-analysis

Portfolio Hiring Analysis

Overview

Parallel deep-dive of every project in a portfolio to map tech stacks, maturity levels, feature gaps, and skill requirements. Synthesizes findings into a hiring strategy with prioritized roles and individual job descriptions ready for posting.

When to Use

• When the user asks "what roles do we need?" or "who should we hire?"
• When planning team growth or restructuring
• When auditing skill coverage across a multi-project portfolio
• When creating job descriptions grounded in actual codebase needs

Inputs

• Project registry — from CLAUDE.md or PROJECTS.md (project names, stacks, domains, status)
• Project directories — actual codebases on disk to analyze

Process

Step 1: Group Projects by Domain

Read the project registry and group active projects into logical domains (e.g., by business area, tech cluster, or team). Skip stale/archived projects unless explicitly requested.

Step 2: Launch Parallel Exploration Agents

Spin up one Agent (subagent_type: Explore, thoroughness: medium) per domain group. Each agent analyzes every project in its group by examining:

• package.json / requirements.txt — dependencies, scripts, package manager
• README.md / CLAUDE.md — project purpose, status, conventions
• src/ or app/ structure — key modules, pages, components, server routes
• Database schema — prisma/schema.prisma, drizzle/, migrations
• CI/CD configs — .github/workflows/, netlify.toml, vercel.json
• Test coverage — test files, feature files, test infrastructure
• Docker setup — Dockerfile, docker-compose.yml
• LLM/AI integrations — LangChain, OpenAI, Anthropic SDK usage
• Content/marketing assets — docs, design files, brand guidelines

Each agent returns a structured summary per project:

## Project Name
**Status:** Active/PoC/MVP/Production/Stale
**Tech Stack:** Framework | ORM | DB | Deploy target
**Codebase Size:** ~N files, N pages, N components
**Key Features:** bullet list
**What's Missing/Incomplete:** bullet list
**Complexity Level:** LOW / MEDIUM / HIGH
**Skills Needed:** categorized by role type

Plus a cross-project summary table and skill requirements grouped by role.

Step 3: Synthesize Hiring Recommendations

Once all agents report back, synthesize into:

1. The Problem — quantify the unsustainability (project count, LOC, framework count, test coverage, CI/CD gaps)
2. Recommended Roles — prioritized table with:
   • Role name
   • Priority (HIGH / MEDIUM-HIGH / MEDIUM)
   • Which projects it covers
   • Type (full-time / part-time / contract / freelance / commission)
3. What You Don't Need (Yet) — roles explicitly excluded with reasoning
4. Team Structure — core team vs extended team suggestion

Step 4: Save Hiring Strategy

Save to ikigai/hiring/strategy.md with:

• Portfolio summary tables (per domain)
• Key findings (patterns, gaps, risks)
• Hiring plan (core + extended team)
• References to individual job descriptions

Step 5: Create Individual Job Descriptions

One file per role in ikigai/hiring/<role-slug>.md. Each JD contains:

# Role Title

**Priority:** HIGH / MEDIUM-HIGH / MEDIUM
**Type:** Full-time / Part-time / Contract / Freelance / Commission
**Location:** Remote / On-site / Hybrid

## Why This Role
1-2 paragraphs connecting role to portfolio needs.

## What You'll Work On
Bullet list: project name (stack) — specific scope.

## Required Skills
Bullet list with experience level hints.

## Nice to Have
Bullet list of bonus skills.

## Day-to-Day
What a typical week looks like.

## Metrics
How success is measured.

## Deliverables (for contract roles)
Numbered list with timelines.

## Compensation Model (for commission roles)
Terms and renegotiation timeline.

Step 6: Present Summary

Show the user a concise priority table and ask if they want to:

• Adjust priorities or scope
• hire agent X to create an AI agent for any role
• Create budget/engagement model breakdown

Output Files

ikigai/hiring/
  strategy.md                  # Portfolio analysis + hiring plan
  <role-slug-1>.md             # Job description per role
  <role-slug-2>.md
  ...

Key Principles

• Evidence-based: Every recommendation traces back to actual codebase analysis, not assumptions
• Parallel execution: Use sub-agents per domain to maximize speed
• Prioritized: Roles ranked by project coverage and revenue impact
• Actionable: Job descriptions ready to post, not abstract wishlists
• Minimal team: Prefer fewer roles covering more projects over specialized silos