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

Claude Model Router

ASecurity

Use when routing Claude models by task — Sonet daily, Opus/Fable complex, mid-task swap, per-task cost tracking. Triggers on "claude model router", "roteamento modelos claude", "sonet vs opus", "swap model mid task", "cost tracking per task", "model routing strategy".

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

Works with

claude codecursorcliapi

Security Analysis

A100/100

Scanned 9/19/2026

Install to Claude Code

$npx -y skills add majinmagros/magros.ai-skills --skill claude-model-router --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Claude Model Router?

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

Security grade badge for Claude Model Router
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/majinmagros-claude-model-router/badge)](https://www.skillsdirectory.com/skills/majinmagros-claude-model-router)

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

Download Zip
Files
SKILL.md
---
name: claude-model-router
description: Use when routing Claude models by task — Sonet daily, Opus/Fable complex, mid-task swap, per-task cost tracking. Triggers on "claude model router", "roteamento modelos claude", "sonet vs opus", "swap model mid task", "cost tracking per task", "model routing strategy".
metadata:
  origin: AUTORAL
  source_docs:
    - https://www.youtube.com/watch?v=Bezlzmti6_U (Luciana Papini video)
    - https://docs.anthropic.com/en/docs/claude-code/settings
  platforms: [claude-code, opencode, cursor, codex, gemini-cli, hermes, openclaw]
  requires_adapters: [hooks, commands]
---

# Claude Model Router — Roteamento por Task

**Sonet** (diário, ≤$0.50/task) · **Opus** (complexo, ≤$5) · **Fable** (reasoning, ≤$10) + swap mid-task + cost tracking. Regras, router e tracker em `references/routing.md`.

## Quando usar (gatilhos concretos)

- "Roteamento de modelos no Claude Code"
- "Quando usar Sonet vs Opus vs Fable"
- "Swap de modelo mid-task"
- "Cost tracking por task"
- "Model routing strategy Claude"

## Quando NÃO usar

- Roteamento geral de LLMs → use `roteamento-modelos-baratos`
- Cost tracking geral → use `cost-aware-llm-pipeline`
- Model routing para outros provedores → use `roteamento-modelos-baratos`

## Regras (resumo)

- Upgrade → Opus se complexidade ≥8/10; downgrade → Sonet se ≤3/10 ou custo >3x esperado; considere Fable se contexto >100k
- `should_swap_model()` + `estimate_cost()` (70/30 in/out) + `record_task()` com budget + `get_session_report()` → `references/routing.md`
- CLI: `claude-model-router swap --to opus-5 --reason "..."` · `status` (modelo, custo, budget)

## Validação Contra Fonte

| Claim | Fonte | Status |
|---|---|---|
| Sonet diário / Opus-Fable complexo / swap economiza | Luciana Papini video | ✅ |
| Cost tracking per task | Anthropic Docs | ✅ |

## Enriquecimento 2026-09-06 — Opus 5 + rate limits (AI Code King `6m1vJqdsanQ`, docs oficiais)

- **Opus 5**: `claude-opus-5`, contexto **1M default e máximo**, 128K output, thinking on — [what's new](https://platform.claude.com/docs/en/about-claude/models/whats-new-opus-5). Preço API **$5/$25** por MTok.
- **Correção**: os **$10/$50** citados no vídeo são do **Fast mode / era Opus 4.8** (Week 30 digest), não do Opus 5 base. Não orçar Opus 5 a $10/$50.
- **Rate limits (14/set)**: boost temporário de 50% expira 13/set; aumento permanente de 25% sobre o baseline = **-17% vs o que você tem hoje** (admitido pela Anthropic). Limites 5h dobrados (maio) permanecem. Auto-mode classifier calls não contam mais no uso; auto-continue no reset reduz babysitting.
- Regra de roteamento: Opus 5 1M não dispensa curadoria — 1M cheio de ruído perde para 200K limpo (retrieval MRCR v2 cai com volume).

## Enriquecimento 2026-09-13 (Batch 16, #49) — Fable 5.1 na prática

- Teste prático (Campelo `5tSGf1DYKe0`): 1 prompt gera landing Awwwards
  (Three.js + partículas + scroll morph). Esforço máximo = multi-agente
  e "gasta fácil" muitos tokens — não é default para toda task.
- Claim do vídeo (2x mais rápido com metade dos tokens no esforço máximo)
  é medição do autor: valide na sua task antes de orçar por ele.
- Regra: esforço máximo só quando a task exige; no resto, o esforço
  padrão entrega igual por menos tokens.

## Enriquecimento rodada 5 — matriz effort + cache (Claude oficial)

- **Comece em medium, suba se preciso** — esforço default medium; só escale
  para high/max quando a task prova que precisa (qualidade insuficiente ou
  horizonte longo). Esforço máximo de saída não é default.
- **Regra do cache** — em threads agênticas longas, leituras repetidas do
  contexto podem custar ~10% via cache. Prompts estáveis + fluxo repetitivo
  = exija cache; sem cache, troque o fluxo antes de trocar o modelo.
- **Modelo mais inteligente pode custar menos por tarefa** — Fable/Opus que
  resolve em 1 passada sai mais barato que Sonnet em 5 tentativas. Compare
  custo-por-tarefa-concluída, nunca preço-por-token.
- **Subagents sempre baratos** — workers delegados rodam no modelo barato
  por padrão; só o architect/orchestrador principal usa o modelo forte.

## Referências Oficiais

- [Luciana Papini Video](https://www.youtube.com/watch?v=Bezlzmti6_U)
- [Anthropic Model Docs](https://docs.anthropic.com/en/docs/claude-code/settings)
- [What's new in Claude Opus 5](https://platform.claude.com/docs/en/about-claude/models/whats-new-opus-5)
- [Claude Code Week 30 — fast mode Opus 4.8 $10/$50](https://code.claude.com/docs/en/whats-new)

## Adapters (Por Plataforma)

```
adapters/
├── opencode/ (hooks, README)
├── cursor/ (hooks, README)
├── codex/ (hooks, README)
└── ...

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 →