Add M3TA OS agent architecture and EEVL workflow automation system

[EagleEyeVisionLabz]

Summary

This PR introduces M3TA OS, a comprehensive file-tree agent architecture for Eagle Eye Vision Labz (EEVL) built on Jake Van Clief's "Agents as Folders" framework. The system orchestrates multiple autonomous agents, sub-agents, and MCP servers to automate business workflows across client onboarding, media production, content pipeline, and event management.

Key Changes

Core Architecture

• orchestrator.json: Master configuration governing all EEVL workflows, sub-agents, and MCP servers
• CLAUDE.md & CLAUDECLAW.md: Comprehensive documentation of the M3TA OS system design and integration bridge between homelab infrastructure and SaaS tools

Workflow Systems (5 major workflows)

• Client Onboarding: Lead capture → Sales call → Contract/Invoice → Project kickoff (5 tasks with schemas, tools, and prompts)
• Media Production: Pre-production → Shoot day → Post-production → Client delivery (4 tasks with full automation)
• Content Pipeline: Idea harvest → Content creation → Design assets → Multi-platform distribution (4 tasks)
• Club Trapeze Management: Event promotion, social media, revenue reporting, ops strategy (4 tasks)
• Gold Star Flag Football: Registrations, event coverage, season reporting, tryout campaigns (4 tasks)

Sub-Agents (8 autonomous agents)

• Onboarding Agent: Executes complete client kickoff within 5 minutes of contract signing
• SocialBot: Multi-platform content distribution across Facebook, Instagram, TikTok, LinkedIn, email
• MeetingScribe: Processes Lark meeting recordings into structured summaries and ClickUp tasks
• TaskMaster: Intelligent task management and auto-assignment based on team capacity
• DocuMind: Contract analysis and legal document risk flagging
• CodeReview: Automated GitHub PR review and feedback
• EEVL Creative: Brand-kit design generation and asset management
• Sparrow Handoff: Post-production file packaging and notification

MCP Servers & Integrations (12 configured)

• CRM/Data: GoHighLevel, Airtable, Notion, Google Drive
• Project Management: ClickUp with AI automation
• Communication: Lark (messaging + meetings)
• Payments: Stripe (invoicing, subscriptions)
• Content/Design: Canva API, OpenRouter (multi-model LLM routing)
• Web Scraping: Apify (TikTok, Instagram, trend monitoring)
• Automation Engines: n8n, Make.com, Lark Bot, GHL workflows

Knowledge Base & Skills

• MyMind AI: Second brain integration with 5 Claude Code skills (/vault-setup, /pdf-synthesize, /tlddr, /canvas-gen, /daily-brief)
• Auto Research: Self-improving skill optimization system with eval suites for continuous improvement

Configuration Files

• Each workflow includes: schema.json (input/output specs), tools.json (API/MCP definitions), prompts.md (agent instructions), and workflow.json (orchestration)
• Each sub-agent includes: config.json (trigger/model settings), data/schema.json, tools.json, and prompts.md

Implementation Details

• Schema-driven: All tasks use JSON Schema for input/output validation and type safety
• Prompt-based: Each agent has detailed role definitions, capabilities, and execution instructions
• Tool abstraction: Unified tool definitions across MCP servers, APIs, and automation engines
• Trigger-based: Sub-agents activate on specific events (contract_signed, lark_meeting_end, github_pr_opened, etc.)
• Modular design: Workflows composed of independent tasks that can be chained or executed standalone
• Multi-model support: OpenRouter integration enables routing to Claude, GPT-4o, Gemini, NVIDIA Nematron based on task requirements

Files Added

• 100+ configuration and schema files
• 40+ prompt files defining agent behaviors
• 30+ tool definitions
• 5+ workflow orchestration files
• 8+ sub

https://claude.ai/code/session_01Q5hzLMKoYx5NDreQZdy9cC

[TechHutTV]

@EagleEyeVisionLabz what the hell is this