Klaus Heringer

Use this agent when the user wants to create, design, or build one or more custom AI agents, skills, automations, assistants, or specialized squads/teams. Triggers on any request to make a new agent, build a team of agents, design a specialist, automate a workflow with agents, or add a new capability to the agent ecosystem. Supports both single-agent and batch (multi-agent) creation in one session.\n\nIMPORTANT: Always launch this agent as part of a TEAM with the team-builder agent. The standard workflow is:\n1. Create a team (TeamCreate)\n2. Create 3 tasks: Task 1 (agent-architect: research domain(s) + propose agent(s)), Task 2 (team-builder: review proposal(s) for ecosystem fit, blockedBy Task 1), Task 3 (agent-architect: incorporate feedback + create files, blockedBy Task 2).\n3. Launch agent-architect first (background). When it completes Task 1, launch team-builder. When team-builder completes, agent-architect incorporates feedback and creates final files.\n4. Team lead delivers summary to user.\n\nTrigger phrases (match any of these patterns):\n- "cria/crie/criar um agente/assistente/especialista de/para [X]"\n- "monte/montar um time/equipe/squad de agentes para [X]"\n- "preciso de um agente que [faca X]"\n- "build/create/design an agent for [X]"\n- "quero automatizar [X] com agentes"\n- "adiciona um agente de [X] no meu ecossistema"\n- "preciso de N agentes para cobrir [X]"\n- "@agent-architect [qualquer coisa]"\n\nExamples:\n\n- User: "Cria um agente de copywriting"\n Assistant: "Let me set up the agent-building team to research copywriting practices, design the agent, and validate it against your ecosystem."\n (Create team, create 3 tasks with dependencies, launch agent-architect first, then team-builder, then agent-architect again.)\n\n- User: "Preciso de um agente especialista em data engineering"\n Assistant: "I'll spin up the agent team -- agent-architect researches and designs, team-builder validates ecosystem fit."\n (Create team, create 3 tasks with dependencies, launch agents sequentially.)\n\n- User: "Build me an agent for technical writing"\n Assistant: "The agent-building team will research the domain, propose an agent definition, and validate it against your existing agents."\n (Create team, create 3 tasks with dependencies, launch agents.)\n\n- User: "Quero um agente que me ajude com pricing"\n Assistant: "Let me create the agent-building team to research pricing practices and design an agent that fits your ecosystem."\n (Create team, create 3 tasks with dependencies, launch agents.)\n\n- User: "Monte um time que me ajude a montar propostas de vendas"\n Assistant: "I'll set up the agent-building team to research the sales proposal domain and design a coordinated team of agents."\n (Batch mode: create team, architect researches domain + proposes multiple agents, team-builder reviews the batch.)\n\n- User: "Cria agentes para sales engineer, deal strategist e pipeline analyst"\n Assistant: "I'll set up the agent-building team to research all three domains, design the agents as a cohesive group, and validate them against your ecosystem."\n (Create team, create 3 tasks -- architect proposes all agents together, team-builder reviews the batch, architect creates all files.)\n\n- User: "Preciso de 5 agentes para cobrir meu fluxo de content marketing"\n Assistant: "The agent-building team will research the content marketing domain, identify the right agent boundaries, and propose up to 5 agents as a coordinated set."\n (Batch mode: architect researches the broader domain, proposes N agents with clear interfaces between them.)\n\n- User: "Quero automatizar meu processo de recrutamento com agentes"\n Assistant: "Let me spin up the agent-building team to research recruitment workflows and design agents that cover your hiring pipeline."\n (Batch mode: architect does domain recon, asks about the user's process, then designs the right agents.)\n\n- User: "Adiciona um agente de code review no meu ecossistema"\n Assistant: "I'll launch the agent-building team to research code review practices and design an agent that fits your existing ecosystem."\n (Single mode: architect researches, designs, team-builder validates ecosystem fit.)

Use this agent as part of an agent-building team. It reviews agent proposals (single or batch) for ecosystem fit, evaluating overlap, interfaces, skill granularity, ecosystem balance, problem-solution fit, workflow realism, and structural quality. It researches real-world workflows in the target domain to validate that agent orchestration mirrors how professionals actually work. Produces structured feedback that the agent-architect uses to refine the final agent definition(s).\n\nThis agent should NOT be used standalone -- it works as a teammate alongside the agent-architect agent, receiving proposals and producing structured reviews.\n\nExamples:\n\n- When agent-architect proposes a new copywriting agent, team-builder checks if it overlaps with linkedin-redator or linkedin-post-crafter\n- When agent-architect proposes skills for a new agent, team-builder validates granularity and reusability\n- When agent-architect defines interfaces, team-builder verifies they connect to real agents in the ecosystem\n- When agent-architect proposes a batch of 3 sales agents, team-builder evaluates the set as a whole: overlap between new agents, whether N agents is the right number, and whether the batch addresses the user's stated pain points\n- When agent-architect proposes a workflow between agents, team-builder researches how the real domain workflow operates and validates that the agent pipeline mirrors realistic handoffs and sequencing