Meta Ads Super Manager Skill

Orquestra planejamento, operacao e otimizacao de campanhas Meta Ads usando MCP tools com seguranca de escrita, cobertura completa de endpoints e carregamento...

Security Scan

The skill is a coherent Meta Ads operations playbook, but it expects to perform read/write actions against MCP/Meta endpoints while declaring no credentials or install requirements — that mismatch is suspicious and needs clarification before trust.

Purpose & Capability

Name, description and the SKILL.md consistently document a Meta Ads/MCP tooling playbook; the listed MCP tool names and workflows align with the stated purpose. However, the skill describes writing actions (create/update/delete) against ad accounts but declares no primary credential, API token, or required environment variables — this omission is unexpected for a tool that performs writes.

Instruction Scope

Runtime instructions are self-contained and restrict accesses to files in the skill package (references/, workflows/, checklists/). They do not direct the agent to read arbitrary host files, external paste links, or unknown network endpoints beyond the MCP tool names. Safety rules and checklists explicitly require user confirmation for destructive actions.

Install Mechanism

Instruction-only skill with no install spec and no code to write to disk; low install risk. All logic is delivered as documentation files, so there are no embedded downloads or archive extraction concerns.

Credentials

The skill expects interaction with MCP/Meta endpoints including writes and uploads, but requires.env and primary credential are empty. A legitimate implementation would typically require one or more tokens/credentials (MCP server token, Meta access token, or platform connector). The absence of declared credentials is disproportionate to its stated capability and creates ambiguity about where/how the skill will obtain permissions.

Persistence & Privilege

Skill is not marked always:true and does not request persistent system modifications. It does instruct operational writes within the ad platform domain but includes checklists and safety guidance; no evidence it modifies other skills or system-wide agent settings.

What to consider before installing

This package is a detailed operational playbook for Meta Ads and appears internally consistent, but it expects to perform read/write operations against MCP/Meta endpoints while declaring no credentials or install steps. Before installing or enabling it: (1) confirm how your agent/environment provides MCP/Meta credentials (and ensure those tokens are least-privilege and user-controlled); (2) verify that the agent will require explicit, interactive confirmation before any destructive writes (deletes, big budget changes); (3) test the workflows in a sandbox ad account or with PAUSED objects only; (4) ask the skill author to list the exact environment variables or connector requirements (e.g., MCP_SERVER_URL, MCP_API_KEY, META_ACCESS_TOKEN) so you can audit credential scope. If the author cannot justify where credentials come from, treat the omission as a risk.

Facebook Ads Manager

Objetivo

Executar campanhas Meta Ads ponta a ponta com disciplina operacional:

• planejar,
• criar estruturas,
• diagnosticar performance,
• escalar com regras matematicas,
• operar audiencias e criativos,
• sempre com controle de risco.

Como usar esta skill sem inflar contexto

1. Use este SKILL.md como roteador.
2. Carregue {baseDir}/references/tools-index.md para escolher tools exatas.
3. Carregue apenas a referencia adicional necessaria para a tarefa:
   • targeting -> {baseDir}/references/targeting-reference.md
   • diagnostico -> {baseDir}/references/diagnostics-reference.md
   • escala -> {baseDir}/references/scaling-rules.md
   • criativos -> {baseDir}/references/creative-frameworks.md
   • seguranca -> {baseDir}/references/safety-policy.md
4. Siga um workflow em {baseDir}/workflows/.
5. Execute checklists em {baseDir}/checklists/ antes e depois de escrita.

Regras nao negociaveis

• Criacoes devem usar status PAUSED por padrao.
• Nunca executar delete sem confirmacao explicita do usuario.
• Mudanca de budget agressiva e proibida sem validacao de risco.
• Nao inventar nome de tool: usar nomes exatos de {baseDir}/references/tools-index.md.
• Sempre tratar act_id no formato act_<id>.
• Em listas paginadas, seguir paging.next com fetch_pagination_url.

Roteamento rapido por intencao

• Planejar conta e baseline -> {baseDir}/workflows/01-plan-and-launch.md
• Diagnosticar queda ou baixa entrega -> {baseDir}/workflows/02-diagnose-and-optimize.md
• Trabalhar CRM, custom e lookalike -> {baseDir}/workflows/03-audience-crm-and-lookalike.md
• Rotina diaria de operacao e escala -> {baseDir}/workflows/04-daily-operations.md

Contrato de resposta

Para tarefas operacionais, responder com:

1. Objetivo e hipotese.
2. Plano de execucao com tools exatas.
3. Acoes feitas e IDs impactados.
4. Proximos passos e monitoramento.
5. Risco e rollback quando houver escrita.

Anti-patterns

• Nao carregar todas as referencias de uma vez.
• Nao usar pseudo-tools (update_x, delete_x).
• Nao ativar campanha/adset/ad sem validacao minima.
• Nao otimizar por metrica unica ignorando funil.