Agents Overview
Agents are the core of EvoNexus. Each agent is a specialized AI persona with its own domain, system prompt, skills, and persistent memory. EvoNexus organizes agents in two ortogonal layers:- Business Layer (17 agents) — operations, finance, community, marketing, HR, legal, product, data, sales, learning retention. Built and maintained by EvoNexus.
- Engineering Layer (21 agents) — software development, code review, testing, debugging, security, design, cycle orchestration, retrospective. 19 derived from oh-my-claudecode (MIT, by Yeachan Heo) + 2 native (Helm, Mirror). Follows a canonical 6-phase workflow. See Engineering Layer for details.
@nova writes a PRD → @apex-architect does the architecture review → @bolt-executor implements → @mirror-retro captures lessons).
Total: 38 agents + custom agents you create.
What Is an Agent?
An agent is a markdown file in.claude/agents/ that contains:
- Frontmatter with metadata (name, description, model, color, memory scope)
- System prompt that defines the agent’s identity, responsibilities, tone, and behavior
How to Invoke an Agent
Slash commands
Each agent has a corresponding command in.claude/commands/:
Business Layer (17):
Auto-routing
You do not need to pick the right agent manually. Just describe what you need and Claude routes to the correct agent based on thedescription field in each agent’s frontmatter.
Examples:
The routing logic uses example interactions in each agent’s
description field to match user intent.
Agent Memory
Every agent has persistent, file-based memory at:
Each memory file uses frontmatter (
name, description, type) and a MEMORY.md index file tracks all entries. Agents read memory at the start of each session and update it as they learn.
Custom Agents
You can create your own agents with thecustom- prefix. Custom agents are gitignored (personal to your workspace) and appear in the dashboard with a gray “custom” badge.
To create a custom agent, use the create-agent skill or see Creating Agents.
Business Layer — 17 Core Agents
Engineering Layer — 19 Agents
Software development specialists. Each agent has a specific role in the dev workflow: design, planning, implementation, review, testing, debugging, verification.Reasoning (opus)
Execution (sonnet)
Speed (haiku)
See the dedicated Engineering Layer page for pipelines, attribution, and detailed workflows.
Agent Roles in Detail
Clawdia is the default agent and operational hub. Morning briefings, email triage, task management, meeting summaries, and end-of-day consolidation all run through Clawdia. Flux acts as a virtual CFO. It queries Stripe for MRR/churn and Omie for payables/receivables, generates financial reports, and manages the monthly close process. Atlas tracks development work across Linear and GitHub. Sprint status, PR reviews, issue tracking, and open source licensing telemetry are its domain. Pulse monitors community health across Discord and WhatsApp. It generates daily/weekly/monthly pulse reports, tracks sentiment, identifies recurring questions, and maintains the FAQ. Pixel manages social media across YouTube, Instagram, and LinkedIn. Content creation, calendar planning, cross-platform analytics, and engagement tracking. Sage handles high-level strategy. OKR reviews, competitive analysis, strategy digests, and decision frameworks. Nex manages the sales pipeline. Lead qualification, proposals, and commercial workflows. Mentor handles the course platform (Evo Academy). Learning paths, module creation, and educational content. Kai is the personal assistant with an isolated domain. Health tracking, habits, personal appointments, and routines. It does not handle professional matters. Oracle is the workspace knowledge agent. It answers questions about how EvoNexus works — agents, skills, routines, integrations, dashboard, configuration, and architecture — by reading the actual documentation before responding. No RAG or vector DB needed. Mako handles marketing. Campaign planning, content creation, brand reviews, SEO audits, email sequences, and performance reports. Aria manages HR and People Operations. Recruiting pipeline, performance reviews, onboarding checklists, compensation analysis, and org planning. Zara handles customer success. Ticket triage, escalation packaging, customer research, draft responses, and knowledge base articles. Lex manages legal and compliance. Contract review, NDA triage, LGPD compliance checks, risk assessment, and legal briefs. Nova handles product management. Feature specs/PRDs, metrics reviews, roadmap updates, product brainstorming, and stakeholder updates. Dex is the data and BI agent. Data analysis, SQL queries, interactive dashboards, statistical analysis, and data validation. Lumen is the knowledge retention agent. It captures atomic facts from pasted content, runs SM-2 spaced repetition review sessions, generates retrieval-practice quizzes, and reports retention metrics per deck. Use Lumen to lock in what you read — Mentor creates learning content, Lumen makes it stick.Preloaded Skills
Each agent has askills field in its frontmatter that pre-loads domain-specific skills into the agent’s context at startup. This means the agent doesn’t need to discover or load skills during execution — they’re immediately available.