Refactor high-complexity React components in Dify frontend. Use when `pnpm analyze-component...
npx skills add TheSimpleApp/agent-skills --skill "ultimate-onboard"
Install specific skill from multi-skill repository
# Description
Drop into ANY project and fully analyze, clean up, document, and configure for parallel agentic work. The nuclear option for project onboarding. Spawns parallel agents for maximum speed.
# SKILL.md
name: ultimate-onboard
description: Drop into ANY project and fully analyze, clean up, document, and configure for parallel agentic work. The nuclear option for project onboarding. Spawns parallel agents for maximum speed.
license: MIT
metadata:
author: thesimpleapp
version: "1.0"
Ultimate Project Onboarding
The nuclear option. Drop into ANY project and make it production-ready for agentic development.
What This Does
ANY PROJECT (messy, undocumented, unfamiliar)
β
βββββββββββββββββββββ
β ULTIMATE ONBOARD β
β β
β 5 Parallel Agentsβ
β Full Analysis β
β Auto Cleanup β
β Complete Docs β
βββββββββββββββββββββ
β
PRODUCTION-READY (clean, documented, configured)
Phases
Phase 1: Parallel Analysis (5 Agents)
Spawn 5 agents simultaneously to map the entire codebase:
Agent 1: STRUCTURE
βββ Map folder structure
βββ Identify entry points
βββ Trace build pipeline
βββ Output: STRUCTURE.md
Agent 2: PATTERNS
βββ Identify coding patterns
βββ Find inconsistencies
βββ Detect frameworks/libraries
βββ Output: PATTERNS.md
Agent 3: DATA
βββ Map database schema
βββ Trace data flow
βββ Identify API contracts
βββ Output: DATA_FLOW.md
Agent 4: DEPENDENCIES
βββ Audit package.json/pubspec
βββ Find outdated packages
βββ Security vulnerabilities
βββ Output: DEPENDENCIES.md
Agent 5: TECH_DEBT
βββ Find code smells
βββ Identify dead code
βββ Measure complexity
βββ Output: TECH_DEBT.md
Phase 2: Synthesis
Combine all 5 reports into unified understanding:
# Architecture Overview
## Tech Stack
- Frontend: [detected]
- Backend: [detected]
- Database: [detected]
- Infrastructure: [detected]
## Key Patterns
- State management: [pattern]
- API layer: [pattern]
- Component structure: [pattern]
## Critical Files
- Entry: [files]
- Config: [files]
- Core logic: [files]
## Health Score
- Code quality: X/10
- Documentation: X/10
- Test coverage: X/10
- Dependency health: X/10
Phase 3: Parallel Cleanup (Conditional)
If user approves, spawn cleanup agents:
Agent A: FORMATTING
βββ Apply consistent formatting
βββ Fix linting errors
βββ Standardize imports
βββ Non-breaking changes only
Agent B: DEAD_CODE
βββ Remove unused imports
βββ Delete commented code
βββ Remove unreachable code
βββ Safe deletions only
Agent C: NAMING
βββ Fix inconsistent naming
βββ Rename unclear variables
βββ Add missing type annotations
βββ Semantic improvements
Agent D: STRUCTURE
βββ Move misplaced files
βββ Create missing folders
βββ Update imports
βββ Organize by feature
Phase 4: Documentation Generation
Auto-generate all project docs:
CLAUDE.md
βββ Project summary
βββ Key patterns
βββ File locations
βββ Commands
.claude/
βββ rules/
β βββ business-logic.md
β βββ coding-standards.md
β βββ architecture.md
βββ standards/
βββ components.md
βββ testing.md
βββ api.md
docs/
βββ ARCHITECTURE.md
βββ API.md
βββ CONTRIBUTING.md
βββ SETUP.md
Phase 5: Configure for Power Mode
Set up for parallel agentic development:
.claude/
βββ settings.json # Model preferences
βββ rules/
βββ parallel-safe.md # Which files can parallel edit
βββ module-owners.md # Agent assignments
βββ no-touch.md # Critical files to protect
Execution Commands
Full Onboard (All Phases)
/ultimate-onboard
β Runs all 5 phases
β Creates all documentation
β Configures for parallel work
Analysis Only
/ultimate-onboard --analyze-only
β Phase 1 + 2 only
β No changes to codebase
β Outputs reports only
Quick Mode
/ultimate-onboard --quick
β Single agent analysis
β CLAUDE.md only
β Fast for familiar codebases
Output Files
| File | Purpose |
|---|---|
CLAUDE.md |
Primary project context |
.claude/rules/*.md |
Coding rules and standards |
docs/ARCHITECTURE.md |
System architecture |
docs/TECH_DEBT.md |
Technical debt backlog |
docs/API.md |
API documentation |
Parallel Safety Rules
When spawning multiple agents:
SAFE to parallelize:
β Different feature folders
β Independent test files
β Separate documentation files
β Non-overlapping modules
NOT safe:
β Shared utilities
β Config files
β Database migrations
β Core type definitions
Detection Patterns
Framework Detection
React: package.json has "react"
Next.js: next.config.* exists
Vue: package.json has "vue"
Flutter: pubspec.yaml exists
Express: package.json has "express"
Supabase: supabase/ folder exists
Pattern Detection
State: Redux | Zustand | Riverpod | Context
API: REST | GraphQL | tRPC
Style: Tailwind | CSS Modules | Styled
Test: Jest | Vitest | pytest | flutter_test
Example Output: CLAUDE.md
# ProjectName
## Overview
E-commerce platform built with Next.js, Supabase, and Tailwind.
## Quick Commands
- `npm run dev` - Start development
- `npm run build` - Production build
- `npm test` - Run tests
## Architecture
- `/app` - Next.js app router pages
- `/components` - React components
- `/lib` - Utilities and helpers
- `/supabase` - Database migrations
## Key Patterns
- Server components by default
- Client components in `/components/client/`
- Supabase for auth and database
- Zustand for client state
## Database
- Supabase project: [ref]
- Key tables: users, products, orders
- RLS enabled on all tables
## Standards
See `.claude/rules/` for coding standards.
Integration
After onboarding, seamlessly use other skills:
/ultimate-onboard β Project fully understood
β
/sprint-planning β Plan features with full context
β
/parallel-swarm β Execute with multiple agents
β
/code-review β Quality gate
Hardware Optimization
With 64 cores / 128GB RAM:
Phase 1: 5 parallel agents (analysis)
Phase 3: 4 parallel agents (cleanup)
Phase 4: 3 parallel agents (docs)
Total time: ~5 minutes for medium codebase
vs ~25 minutes sequential
Rollback
All changes are tracked:
# Undo cleanup changes
git checkout -- .
# Keep only documentation
git checkout -- src/
git add docs/ CLAUDE.md .claude/
# Supported AI Coding Agents
This skill is compatible with the SKILL.md standard and works with all major AI coding agents:
Learn more about the SKILL.md standard and how to use these skills with your preferred AI coding agent.