convez.ai
SkillProdutividadeIntermediário
hhhyz544

Claude Code Focus Router

hhhyz544/claude-code-focus-router

Roteamento de atenção task-first: priorize tarefas reais sobre tom emocional

Ver no GitHub
InicialPluginScore · de 0 a 100

Skill que reorienta a atenção do Claude para tarefas reais, constraints e outputs acionáveis, minimizando ponderação excessiva de tom emocional. Ideal para debugging, troubleshooting e contextos de alta urgência onde clareza técnica importa mais que reconforto.

Instalação

$npx convez-ai hhhyz544/claude-code-focus-router

1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.

Métricas

Estrelas
1
Forks
0
Acompanhando
0
Saúde
57%
Último commit
há 6 meses
Versão
v0.1.1
Issues abertas
0
Contribuidores
2

Casos de uso

  • Debugging e troubleshooting sob pressão ou frustração
  • Requisições de respostas diretas e sem fluff
  • Tarefas de risco elevado onde limites de segurança e próximos passos importam
  • Design de prompts e workflows para roteamento de atenção mais rigoroso

Veredito

Skill útil e bem-construída para refinamento de qualidade de resposta em contextos técnicos e urgentes. Abordagem pragmática que reconhece limites reais da calibração emocional em IA sem ser manipuladora.

README

A tradução para o português deste README ainda está na fila.

Claude code focus-router

Claude code focus-router is a Claude Skill for task-first attention routing. It helps Claude prioritize the user's real task, constraints, risks, evidence, and actionable output, while using emotional tone only to calibrate communication style.

This repository contains a Claude Skill. It is not an independent AI model, a prompt-injection filter, or a multi-agent framework.

Problem

AI assistants can sometimes over-weight a user's emotional tone and under-weight the actual task. That can lead to responses that are polite but less useful: too much reassurance, too many apologies, unnecessary clarification questions, or missing technical details.

FocusRouter gives Claude a compact set of behavioral instructions for keeping attention on:

  • The user's primary task
  • Required output format
  • Constraints and risks
  • Evidence and uncertainty
  • Actionable next steps
  • Tone calibration as a secondary concern

Use Cases

Use this Skill when you want Claude to stay task-focused in situations such as:

  • Debugging or technical troubleshooting under frustration or urgency
  • Requests like "be direct", "no fluff", "only give steps", or "do not comfort me"
  • High-risk topics where safety boundaries and next steps matter
  • Ambiguous requests where Claude should state reasonable assumptions and continue
  • Prompt or workflow design for more rigorous AI attention routing
  • Review, critique, or planning tasks where emotional tone should not replace analysis

Installation

Project-level install

Copy the Skill folder into the target project's .claude/skills/ directory:

mkdir -p /path/to/project/.claude/skills
cp -R .claude/skills/focus-router /path/to/project/.claude/skills/

Then start Claude Code from that project:

cd /path/to/project
claude

User-level install

To make the Skill available across your Claude Code projects:

mkdir -p ~/.claude/skills
cp -R .claude/skills/focus-router ~/.claude/skills/

If Claude Code was already running before the skills directory existed, restart the session so Claude Code can discover it.

Usage

Claude can load the Skill automatically when your request matches the description in SKILL.md.

You can also invoke it directly:

/focus-router I am in a hurry. Do not comfort me. Give me the debugging steps for a Node EADDRINUSE error.

Compatibility

This repository uses the Claude Skill folder structure:

.claude/skills/focus-router/SKILL.md

It is intended for Claude Code environments that support project-level or user-level Skills. It is not an npm package, Python package, standalone model, browser extension, or multi-agent framework.

No scripts, hooks, tool permissions, or automatic commands are required for this Skill.

Security

This repository contains a text-only Claude Skill. It does not include executable scripts, package dependencies, GitHub Actions workflows, or requested tool permissions.

Security policy and reporting guidance are in SECURITY.md.

Example Trigger Phrases

  • "Do not comfort me; solve it directly."
  • "Give me a more rational answer."
  • "No fluff."
  • "Only give steps."
  • "I am in a hurry; tell me what to do."
  • "You did not answer the core question."
  • "Help me write a prompt that keeps AI focused on the task."
  • "Make Claude avoid being distracted by emotion."
  • "Create an attention-routing skill."

Testing

After installing the Skill, start Claude Code in a project where the Skill is available and try these prompts:

/focus-router I am frustrated. Do not comfort me. Give me the exact steps to debug a Node EADDRINUSE error.

Expected behavior: Claude should prioritize concrete debugging steps, state assumptions briefly, and avoid extended reassurance.

/focus-router I am in a hurry and want to delete generated files. What is the safest way to do it?

Expected behavior: Claude should identify deletion risk, suggest safer verification steps, and avoid vague encouragement.

/focus-router This PR feels terrible. Find the core problems and do not ask a pile of questions.

Expected behavior: Claude should state a reasonable assumption, proceed with a review-oriented approach, and keep the answer actionable.

Project Structure

focus-router/
|-- .gitignore
|-- CHANGELOG.md
|-- README.md
|-- LICENSE
|-- SECURITY.md
`-- .claude/
    `-- skills/
        `-- focus-router/
            `-- SKILL.md

Design Principles

  • Task goals take priority over emotional reaction.
  • Emotion may calibrate tone, but it must not change factual judgment.
  • Safety, risk, and uncertainty should be stated clearly.
  • When action steps are needed, provide action steps first.
  • When information is incomplete but a reasonable assumption is possible, state the assumption and proceed.
  • Do not reduce technical detail just to sound friendly.
  • Do not use empathy as a substitute for analysis.

License

MIT License. See LICENSE.

Ver o original no GitHub

Bom para

Relacionados