Platform Configuration

This page summarizes .claude/CLAUDE.md — the core platform configuration for Mahoosuc OS agents and workflows.

Project Overview

This project is a comprehensive AI agent orchestration system that combines:

• Prompt Engineering Knowledge Base: Best practices from OpenAI, Anthropic, Google
• Meta-Agents: AI agents that create other AI agents
• Zoho ONE Integration: CRM, Mail, and SMS automation
• Multi-Interface System: CLI slash commands, web UI, and Zoho widgets

Core Philosophy

Data Model Enhancement

• ENHANCE the data model by ADDING columns/fields - We can remove them later after systems are functioning
• ENHANCE OVER REMOVING - Always prefer to add functionality rather than remove existing code
• Iterate with expansion - Build comprehensive systems first, optimize second

Zoho ONE Integration Priority

• CRM: All lead, contact, and deal management flows through Zoho CRM
• MAIL: Email campaigns and communications via Zoho Mail
• SMS: SMS campaigns and notifications via Zoho SMS

Slash Commands

Mahoosuc OS includes 120+ slash commands across 45+ categories. Key categories:

• /architect:plan --title "Feature Name" - Create planning session with isolated git worktree
• /architect:review [workflow-id] --wait - Trigger multi-agent review (Security, Performance, Architecture, Testing)
• /devflow:review --title "Feature Name" - Create in-sprint review session with git branch
• /devflow:compound --category pattern --title "..." --impactScore 8 - Capture compound learnings
• pattern - Reusable design or code pattern discovered
• anti-pattern - Pattern to avoid (learned the hard way)
• best-practice - Recommended approach that worked well
• gotcha - Common pitfall or mistake to watch out for
• tip - Helpful hint or optimization trick
• Learnings automatically integrate into CLAUDE.md and DEVB knowledge base
• 30-day TTL with archive option for cleanup
• Tag-based semantic search for retrieval

For the complete reference, see .claude/SLASH_COMMANDS_REFERENCE.md.

Agent Routing Guidelines

When to Use Which Agent

For Prompt Engineering:

• Use prompt-engineering-agent from /meta-prompts/ for creating new prompts
• Use documentation-expert-agent for creating/updating documentation
• Reference /guides/unified-best-practices__claude_sonnet_4.md for best practices For Zoho Integration:

For the full routing logic, consult .claude/CLAUDE.md.

Quality Standards

All Prompts Must:

• Follow unified best practices from /guides/unified-best-practices__claude_sonnet_4.md
• Include clear role definition and expertise areas
• Specify operational context (domain, audience, quality tier)
• Define input processing protocol
• Specify reasoning methodology (CoT, ReAct, etc.)
• Include quality control checklist
• Provide execution protocol

Available Skills

32 skill packages available for agent preloading (18 top-level + 14 Obra Superpowers).

Category	Count	Examples
Original Skills	13	frontend-design, qa-architect, security-auditor
Vercel Skills	5	ai-sdk, vercel-react-best-practices, web-design-guidelines
Obra Superpowers	14	brainstorming, systematic-debugging, test-driven-development

Skills are preloaded in agent frontmatter via the skills: field, or auto-activated by context.

Standards

Project coding standards are in .claude/standards/:

Path	Scope
.claude/standards/global/	Coding style, conventions, error handling, API contracts
.claude/standards/frontend/	Component patterns, accessibility, performance
.claude/standards/backend/	API design, database, security, architecture
.claude/standards/testing/	Unit, integration, E2E testing patterns