Build a system of 5 autonomous agents with Codex: Campaign Manager, Research, Content Strategy, Creative Copy, and Data Analysis working in parallel.
5 agents, roles, workflow
system.md, brand.md, SKILL.md
orchestration, task_manifest
competitors, data, plan
creative, metrics, report
coordination, parallel work, outputs
The system AgenteMktCX uses Codex as an "autonomous work team." 5 specialized agents work in parallel, each with its own skill, output folder, and coordination rules.
A complete marketing campaign (research + strategy + creative + analysis) can be produced by Codex in minutes, with outputs organized into folders — without external tools or freelancers.
| Agent | Role | Output |
|---|---|---|
| Campaign Manager | Orchestrator | task_manifest.json, final report |
| Research Agent | Researcher | /output/research/ |
| Content Strategy | Strategist | /output/strategy/ |
| Creative Copy | Creative writer | /output/copy/ |
| Data Analysis | Data analyst | /output/data/ |
The system organizes everything in shared context files: system.md defines the orchestration rules, brand.md defines the brand identity, and each agent has its own SKILL.md own.
Crie um arquivo system.md que defina: - Funções de cada agente (Campaign Manager, Research, Content Strategy, Creative Copy, Data Analysis) - Estrutura das SKILLs (SKILL.md por agente) - Pastas /output por agente - Regras de orquestração e handoff - Formato do task_manifest.json para coordenação
Crie brand.md com: - nome: [nome da marca] - público: [perfil do cliente ideal] - tom: [ex: profissional, jovem, técnico] - posicionamento: [proposta de valor] - temas: [tópicos de conteúdo] - palavras-chave: [keywords SEO] - canais: [LinkedIn, Instagram, email...] - metas: [leads, brand awareness, vendas]
mkt-system/
├── system.md
├── brand.md
├── agents/
│ ├── research/SKILL.md
│ ├── strategy/SKILL.md
│ ├── copy/SKILL.md
│ └── data/SKILL.md
└── output/
├── research/
├── strategy/
├── copy/
└── data/
The Campaign Manager is the “campaign director” — it reads brand.md, creates task_manifest.json with each agent’s tasks, and, at the end, consolidates all outputs into an executive report.
You are the Campaign Manager. Read brand.md and create task_manifest.json with the tasks for each agent. Then, once you receive their outputs, generate a consolidated final campaign report at /output/campaign-report.md
{
"campaign": "Q3 Lead Generation",
"tasks": [
{
"agent": "research",
"task": "Analise os 3 principais concorrentes",
"output": "output/research/competitors.md",
"status": "pending"
},
{
"agent": "strategy",
"task": "Crie plano de conteúdo para 30 dias",
"output": "output/strategy/content-plan.md",
"status": "pending",
"depends_on": ["research"]
}
]
}
The Research Agent analyzes the market: it looks up competitor data, industry trends, and opportunities. It saves everything organized in /output/research/.
Crie SKILL.md para o Research Agent que: - Analisa os 3 principais concorrentes da marca - Coleta dados do mercado (tamanho, crescimento) - Identifica lacunas e oportunidades - Salva relatório em /output/research/ - Formato: markdown estruturado com fontes
Use the Research data to create a strategic content plan: a 30-day calendar, topics for each channel, and content pillars aligned with brand.md.
Crie SKILL.md para o Content Strategy Agent: - Lê /output/research/ e brand.md - Cria calendário de conteúdo (30 dias) - Define pilares por canal (LinkedIn, IG, email) - Salva em /output/strategy/content-plan.md
Turns the content plan into ready-to-publish assets: posts for each channel, email headlines, and A/B variations to test. Uses the tone defined in brand.md.
Create SKILL.md for the Creative Copy Agent: - Reads brand.md and /output/strategy/content-plan.md - Writes complete posts for LinkedIn and Instagram - Creates 3 subject line variations for email - Tone: according to brand.md - Saves everything to /output/copy/
Analyzes results from previous campaigns, identifies performance patterns, and suggests data-driven optimizations. Delivers KPI reports in /output/data/.
Crie SKILL.md para o Data Analysis Agent: - Analisa dados de campanhas anteriores (CSV/JSON) - Calcula CTR, conversão, CAC, ROAS - Identifica top 3 conteúdos por engajamento - Recomendações de otimização - Salva relatório em /output/data/analysis.md
O task_manifest.json is the team's "task board." The Campaign Manager creates it, the agents update their statuses, and the manifest ensures the workflow is followed (e.g., strategy only runs after research).
codex "Leia task_manifest.json e execute todas as tarefas pendentes. Para tarefas sem dependência, execute em paralelo. Atualize o status de cada tarefa no manifesto e salve os outputs nas pastas corretas. Ao finalizar tudo, gere campaign-report.md com resumo executivo."
Result: Full campaign in /output/ — research, strategy, ready-to-publish posts, data analysis, and an executive report. Everything organized, everything versioned in git.