9.4 KiB
9.4 KiB
InitX - Project Intelligence & Agent Recruitment System
Usage
/initx [OPTIONS]
🎯 Core Function
Analyze project codebase, detect technology stack, and CREATE NEW PROJECT-SPECIFIC AGENTS tailored to your unique requirements. Saves custom agents to .claude/agents/ and updates project CLAUDE.md.
🚀 Command Modes
🎯 Basic Usage
/initx # Smart analysis & agent creation
/initx --preview # Preview mode (no file creation)
/initx --force # Force overwrite existing agents
📊 Advanced Options
--mode=minimal- Create only essential agents--focus=security- Focus on security-specific agents--exclude=ui- Skip UI/design agents--template=mobile- Use mobile project template--model=inherit- Force all agents to use specified model (inherit/sonnet/opus/haiku)
🧠 Intelligence Detection Matrix
Technology Stack Recognition
Frontend: vue (*.vue, nuxt.config.js), react (*.jsx, next.config.js), angular
Backend: fastapi (main.py), spring (pom.xml), go (go.mod), node (express/koa)
Mobile: android (*.kt, AndroidManifest.xml), flutter (pubspec.yaml)
Security: reversing (*.apk, frida scripts), hooking (xposed modules)
Project Complexity Assessment
- Simple: 1-2 custom agents (single stack)
- Medium: 3-5 custom agents (multi-stack)
- Complex: 5-8 custom agents (microservices)
- Enterprise: 8+ custom agents (distributed)
🤖 Agent Creation Algorithm
Phase 1: Core Stack Agents
Creates NEW specialized agents based on detected stack:
vue-{project-name}-developer- Project-specific Vue expertapi-{domain}-specialist- Custom API integration expert{database}-data-architect- Database-specific data expert
Phase 2: Domain-Specific Agents
Creates NEW agents for unique project domains:
{domain}-business-logic-expert- Business rules specialist{integration}-connector-agent- Third-party integration expert{platform}-deployment-specialist- Platform-specific deployment
Phase 3: Quality & Operations Agents
Creates NEW agents for project-specific QA:
{stack}-testing-specialist- Stack-specific testing expert{environment}-ops-engineer- Environment-specific operations
📁 Generated Structure
project_root/
├── .claude/
│ └── agents/ # NEW custom agents
│ ├── vue-ecommerce-developer.md
│ ├── payment-integration-specialist.md
│ ├── postgres-data-architect.md
│ └── aws-deployment-specialist.md
├── CLAUDE.md # Updated with new team
└── .gitignore # Updated exclusions
🏗️ Custom Agent Template
⚠️ LANGUAGE REQUIREMENT: All agent content MUST be generated in ENGLISH.
---
name: {project-domain}-{specialization}-agent
description: Project-specific {specialization} expert for {project_name}
model: inherit # Options: inherit | sonnet | opus | haiku
---
You are the **{Project Domain} {Specialization} Agent** for the {project_name} project.
## STRICT AGENT BOUNDARIES
**ALLOWED ACTIONS:**
- {Domain-specific actions based on project analysis}
- {Tech stack specific implementations}
- {Integration and performance optimizations}
**FORBIDDEN ACTIONS:**
- {Cross-domain work} (delegate to {appropriate_agent})
- Infrastructure concerns (delegate to devops-engineer)
- Security audits (delegate to code-review-expert)
**CORE MISSION:** {Single sentence describing the agent's purpose}
## RESPONSIBILITIES
### 1. {Primary Domain Area}
- {Specific responsibilities based on project analysis}
- {Pattern implementations from detected codebase}
### 2. {Quality & Testing}
- {Testing requirements specific to domain}
- {Performance criteria for project}
### 3. {Collaboration}
- Input from: {upstream agents}
- Output to: {downstream agents}
- Coordinate with: {peer agents}
## TECHNOLOGY STACK
**Primary**: {Detected technologies with versions}
**Integrations**: {Third-party services}
**Constraints**: Work exclusively within {project_domain} of {project_name}
🧠 Intelligent Model Selection System
Model Selection Matrix
Complex Reasoning (opus): Architecture design, security analysis, microservices
Balanced Performance (sonnet): Most development agents, API/database, testing
Simple Fast (haiku): Config generation, simple scripts, documentation
Inherit from Parent (inherit): Maintain consistency, avoid switching overhead
Force Model Option (--model parameter)
/initx --model=inherit # All agents inherit from parent (best for cost control)
/initx --model=sonnet # All agents use Sonnet (balanced performance)
/initx --model=opus # All agents use Opus (maximum reasoning)
/initx --model=haiku # All agents use Haiku (fastest response)
⚠️ Using --model overrides intelligent model selection for consistency/cost control.
User Confirmation Flow
Standard Mode (Intelligent Model Selection)
🤖 Agent Creation - {agent_name}
📊 Recommended: {suggested_model} ({reason})
Select model: [1] ✅ Recommended [2] Opus [3] Sonnet [4] Haiku [5] Inherit
Confirm? [y/N]
Force Model Mode (--model parameter)
🤖 Agent Creation - {agent_name}
📊 Model: {forced_model} (forced by --model)
Confirm? [y/N]
⚠️ IMPORTANT: Language Requirements
All generated agents MUST be created in English, regardless of the user's input language:
- Agent names: Always in English (e.g.,
vue-ecommerce-developer, notvue-电商-开发者) - Agent descriptions: Always in English
- Agent content: Always in English
- Documentation comments: Always in English
This ensures consistency and compatibility across all projects and teams.
⚡ Execution Flow
Phase 1: Project Intelligence (15s)
🔍 Analyzing project structure...
├── 📊 Detecting tech stack and versions
├── 🏗️ Identifying architecture patterns
├── 📈 Assessing complexity and scale
└── 🎯 Discovering unique requirements
✅ Analysis Complete
- Type: E-commerce Web App
- Stack: Vue 3 + FastAPI + PostgreSQL
- Complexity: Medium (4/5)
- Unique: Payment integration, inventory management
Phase 2: Agent Creation & Confirmation (20s)
Standard Mode:
🤖 Intelligent Model Matching...
vue-ecommerce-developer (recommended: sonnet)
├─ Reason: Vue 3 patterns, balanced performance
├─ Select: [1] ✅ sonnet [2] opus [3] haiku [4] inherit
└─ ✅ Created
fastapi-payment-specialist (recommended: opus)
├─ Reason: Payment integration, complex logic
├─ Select: [1] ✅ opus [2] sonnet [3] haiku [4] inherit
└─ ✅ Created
[... more agents ...]
📊 Created 8 agents: 4 sonnet, 2 opus, 1 haiku, 1 inherit
Force Model Mode (/initx --model=inherit):
🤖 Force Model: inherit (all agents use parent model)
vue-ecommerce-developer
├─ ⚠️ Using inherit (recommended: sonnet)
└─ ✅ Created
fastapi-payment-specialist
├─ ⚠️ Using inherit (recommended: opus)
└─ ✅ Created
[... more agents ...]
📊 Created 8 agents: 8 inherit (forced)
Phase 3: Team Configuration (10s)
📝 Configuring AI team collaboration...
├── 📁 Saving agents to .claude/agents/ directory
├── 🔧 Setting up collaboration matrix and dependencies
├── 📋 Updating CLAUDE.md with team information
└── 🎯 Ready for project-specific task execution
✅ Custom AI team ready!
🚀 Use /ai "optimize checkout flow" to engage specialist team
🎛️ Smart Templates
E-commerce Project
detected_patterns: [shopping_cart, payment_flow, inventory]
created_agents:
- payment-integration-specialist
- inventory-management-expert
- customer-analytics-agent
- ecommerce-testing-specialist
Security Research Project
detected_patterns: [apk_analysis, hooking, reverse_engineering]
created_agents:
- android-malware-analyst
- frida-hooking-specialist
- apk-reverse-engineer
- vulnerability-scanner-expert
Microservices Project
detected_patterns: [service_mesh, api_gateway, distributed_data]
created_agents:
- service-mesh-architect
- api-gateway-specialist
- distributed-data-expert
- microservices-testing-specialist
🚀 System Benefits
🎯 Project-Specific Intelligence
- Custom Agents: Tailored to your exact tech stack and domain
- Smart Collaboration: Agents know your project's integration points
- Model Flexibility: Choose optimal model strategy with
--modelparameter
⚡ Rapid Development
- 10x Faster: Custom experts vs generic agents
- Zero Learning Curve: Agents pre-configured for your project
- Cost Control: Force consistent model usage with
--model=inherit
💰 Cost Management
- Default: Intelligent per-agent model selection
- --model=inherit: Maximum cost control
- --model=sonnet: Balanced cost/performance
- --model=haiku: Minimum cost, fast response
InitX creates your perfect AI development team - not just assigns existing ones!
🚀 One command, custom specialists, exponential productivity!