2026-02-01 project open-source claude-code plugin skills agents

Solo Factory

From shower thought to deployed product — 30 skills, 3 agents, and a code intelligence MCP server that knows every project you’ve ever built.

/research → /validate → /scaffold → /setup → /plan → /build → /deploy → /launch → /review

Why

Most AI coding tools help you write code. Solo Factory helps you run a startup — from market research and idea validation to deployment and promotion. Every skill is designed for one-person teams moving fast across multiple projects and stacks.

No context switching — one pipeline handles research, coding, deployment, and marketing
Stack-agnostic — 9 templates from SwiftUI to Cloudflare Workers
Code intelligence — Solograph MCP server indexes all your projects, searches past sessions, and provides semantic code search
Works without MCP — skills gracefully fall back to Glob, Grep, Read, WebSearch

Install

# Option 1: Skills (any AI agent — Claude Code, Cursor, Copilot, Gemini CLI, Codex)
npx skills add fortunto2/solo-factory --all

# Option 2: Claude Code Plugin (skills + agents + MCP)
claude plugin marketplace add https://github.com/fortunto2/solo-factory
claude plugin install solo@solo --scope user

Skills

30 executable skills in 4 phases: Analysis (6) → Development (6) → Promotion (11) → Utility (7).

Browse all skills → — descriptions, install commands, full SKILL.md source for each.

Agents

Agent	Model	Specialization
`researcher`	Sonnet	Market research, competitors, pain points
`code-analyst`	Haiku	Codebase exploration, dependency analysis
`idea-validator`	Sonnet	Idea validation, STREAM scoring, PRD generation

Stacks

9 stack templates: iOS, Next.js, Kotlin, Astro, Python, Cloudflare Workers.

Browse all stacks → — full YAML configs with copy button.

Automated Pipeline

Two modes for hands-free execution:

# Interactive — one Claude Code session
/pipeline dev "my-app" "nextjs-supabase"

# Big Head — tmux dashboard, logs, long pipelines
make bighead-dev P=my-app S=nextjs-supabase

Pipeline signals: <solo:done/> (stage complete) and <solo:redo/> (go back). Stage markers in .solo/states/. Per-iteration logs in ~/.solo/pipelines/.

MCP Integration

Skills auto-detect Solograph tools when available:

Tool	What it does
`kb_search`	Semantic search over knowledge base
`session_search`	Search past Claude Code sessions
`codegraph_query`	Cypher queries on code intelligence graph
`codegraph_explain`	Architecture overview of any project
`project_code_search`	Semantic code search across projects
`web_search`	Web search via SearXNG or Tavily

Without MCP, skills fall back to Glob, Grep, Read, WebSearch — always works.

Architecture

The harness IS the product:

skills/          — 30 SKILL.md files (instructions, not code)
agents/          — 3 agent definitions (model + role + tools)
templates/stacks — 9 YAML stack templates
rules/           — routing table, AI comments, debugging
hooks/           — session start/stop, pipeline progression
scripts/         — bighead launcher, tmux dashboard, formatters

Skills are prompts, not programs. Agent mistake → fix the skill, not the code.