Skills DirectorySkills Directory
SkillsLearnSecurityCategoriesDocsCommunityBlog
Sign InSubmit Skill
Skills Directory

Security-tested agent skills for Claude, coding agents, and AI workflows.

Directory

  • Browse Skills
  • All Skills A–Z
  • Claude Skills
  • Claude Code Skills
  • Agent Skills
  • Categories
  • Submit a Skill

Learn

  • Learn Hub
  • Install Claude Skills
  • Write SKILL.md
  • Skills vs MCP
  • Directories Compared

Security

  • Security
  • Methodology
  • Secure Claude Skills
  • Security Badges

Company

  • About
  • Community
  • Blog
  • API Docs
  • Advertise

2026 Skills Directory. All rights reserved.

Back to skills

Attention Architecture

ASecurity

Supervise N agents with 1 brain during-run. Use when coordinating 3+ parallel agents, reviewer is overloaded, routine approvals bury exceptions. TRIGGER when: attention bottleneck, exception inbox, batch approvals, escalation path, N:1 review ratio. Gatilhos PT: arquitetura de atenção, inbox de exceções, revisar N agentes, gargalo de supervisão, escalação para humano.

2 stars
0 votes
0 copies
0 views
Added 9/19/2026
ai-agentsgorails

Security Analysis

A100/100

Scanned 9/19/2026

Install to Claude Code

$npx -y skills add majinmagros/magros.ai-skills --skill attention-architecture --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Attention Architecture?

Add the live security badge to your README — it updates automatically with every re-scan.

Security grade badge for Attention Architecture
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/majinmagros-attention-architecture/badge)](https://www.skillsdirectory.com/skills/majinmagros-attention-architecture)

More formats (shields.io, HTML) on the badges page.

Download Zip
Files
SKILL.md
---
name: attention-architecture
description: Supervise N agents with 1 brain during-run. Use when coordinating 3+ parallel agents, reviewer is overloaded, routine approvals bury exceptions. TRIGGER when: attention bottleneck, exception inbox, batch approvals, escalation path, N:1 review ratio. Gatilhos PT: arquitetura de atenção, inbox de exceções, revisar N agentes, gargalo de supervisão, escalação para humano.
---

# Attention Architecture

## Quando usar

- 3+ agentes rodando em paralelo e 1 humano/lead como revisor.
- Revisor afogado em log de rotina e perdendo falha crítica.
- Decisões similares pedindo aprovação uma a uma.
- Precisa definir quando puxar humano vs. deixar agente seguir.
- NÃO é pré-run. Pré-run é `swarm-readiness-gate`. Este skill é DURANTE-run.

## Passos

1. **Padronize o formato de evidência.** Exija de todo agente o mesmo envelope: `decisão | evidência (links/arquivos) | confiança | risco | alternativa descartada`. Sem envelope, sem review.
2. **Separe rotina de exceção.** Crie inbox de exceções: só sobe para humano o que for baixa confiança, alto risco, ou fora do padrão. Rotina vai para log, não para fila humana.
3. **Faça batching de decisões similares.** Agrupe por tipo (ex: 12 migrações de schema, 8 textos de UI). Aprove/rejeite o lote com uma regra, não 12 vezes.
4. **Automatize os checks chatos.** Lints, diff de schema, comparação visual, teste determinístico. Humano só julga o que máquina não julga bem.
5. **Defina escalation path explícito.** Documente: condição de trigger → quem é puxado → SLA → o que o agente faz enquanto espera (para, segue em dry-run, desfaz).
6. **Meça N:1.** Acompanhe `N agentes ativos vs 1 revisor` + tempo médio por exceção + exceções perdidas. Se o revisor virou gargalo, reduza N ou endureça o filtro, nunca "revise mais rápido".

## Regras

- Nunca revise raw log. Só revise envelope de evidência padronizado.
- Exceção primeiro, rotina nunca interrompe exceção.
- Sem escalation path escrito, o agente NÃO roda autônomo.
- Batch sem critério explícito é proibido — vira aprovação cega.
- Métrica N:1 sempre visível. Atenção é recurso finito, trate como orçamento.
- Humano puxado tarde é falha do filtro, não do humano. Ajuste o filtro.

## Related skills

- `swarm-readiness-gate` — pré-run, este é durante-run.
- `agent-swarm-ops` — operação do swarm após o GO.
- `agent-guardrails` — harness engineering contínuo durante o run.

Attribution

majinmagrosmajinmagros
View sourceMore from majinmagros →
SSkills DirectorySkills Directory

Ship a skill? Prove it's safe.

Free 120-pattern security scan, letter grade, and an embeddable README badge.

Submit a skill

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments (0)

No comments yet. Be the first to comment!

SSkills DirectorySkills Directory

Ship a skill? Prove it's safe.

Free 120-pattern security scan, letter grade, and an embeddable README badge.

Submit a skill

Related Skills

Caveman

Ultra-compressed communication mode. Cuts token usage ~75% by speaking like caveman while keeping full technical accuracy. Supports intensity levels: lite, full (default), ultra, wenyan-lite, wenyan-full, wenyan-ultra. Use when user says "caveman mode", "talk like caveman", "use caveman", "less tokens", "be brief", or invokes /caveman. Also auto-triggers when token efficiency is requested.

1023331 votes

Hyperplan

Adversarial multi-agent planning skill. Self-orchestrates 5 hostile category members (unspecified-low, unspecified-high, deep, ultrabrain, artistry) via team-mode for ruthless cross-critique debate, distills only the defensible insights, then MANDATORILY hands the distilled insight bundle to the `plan` agent for executable plan formalization. Use when planning needs maximum rigor and surfacing of weak assumptions, blind spots, and over-engineering. Triggers: 'hyperplan', 'hpp', '/hyperplan', ...

686011 votes

Mcp Code Execution

Routes multi-tool workflows through MCP servers for large datasets and pipelines. Use when Bash tool overhead is limiting throughput on data-heavy tasks.

3331 votes

catchup

Recovers prior coding-agent session context by running `catchup <agent> --since-compact`, which extracts a clean summary of a previous Codex, Claude Code, Antigravity, OpenCode, or Pi Agent session. Use when the user says "catch up", "what did the last session do", "get me up to speed", "I switched agents", or asks to recover/summarize a previous session before continuing. Do NOT use for the current conversation, git history, or any non-agent log.

611 votes

math-skill

A comprehensive mathematical reasoning skill for AI assistants — handles arithmetic to research-level problems with rigorous step-by-step reasoning, systematic verification, and transparent uncertainty handling

381 votes
View all in ai-agents →