convez.ai
SkillProdutividadeIntermediário
AI-Business-Tools

Claude Code Skills Library

AI-Business-Tools/claude-code

Skills estruturados para transformar prompting ad-hoc em workflows profissionais repetíveis

Ver no GitHub
BomPluginScore · de 0 a 100

Biblioteca de skills reutilizáveis para Claude Code focada em workflows profissionais: geração de slides, gestão de conhecimento, análise de documentos e escrita com voz consistente. Desenvolvida originalmente para materiais de curso MBA, generalizada para qualquer profissional de conhecimento que produz documentos, apresentações e análises recorrentes.

Instalação

$npx convez-ai AI-Business-Tools/claude-code

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
7
Forks
1
Acompanhando
1
Saúde
50%
Último commit
há 3 semanas
Versão
—
Issues abertas
0
Contribuidores
3

Casos de uso

  • Geração automática de decks de slides a partir de conteúdo acadêmico ou corporativo
  • Gestão estruturada de conhecimento com rastreamento de versões e handoffs
  • Análise consistente de documentos com padrões de qualidade definidos
  • Produção de materiais de curso (exercícios, cases, sumários de discussão)
  • Criação de relatórios e propostas com voice profissional padronizado

Veredito

Abordagem madura e bem documentada para elevar Claude Code além de prompting ad-hoc. Particularmente valioso para quem produz documentos recorrentes em volume, codificando padrões de qualidade e preferências que persistem entre sessões. Execução clara com filosofia sólida de camadas (regras, reforço, execução, memória).

README

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

Claude Code

A getting-started guide and a skills library for business professionals and educators.

Start here

New to this repository? Work through it in order:

  1. Get started. What Claude Code adds beyond Claude chat and Cowork, how to install it on Mac or Windows, and how to set up your first project.
  2. Set up version history. Put your Claude Code environment under Git and sync two machines through a private GitHub repository, so every change to your skills and standards is tracked and reversible.
  3. Adopt handoffs and resume. End each session with a structured summary so the next one picks up where you left off.
  4. Browse the skills. Reusable workflows for slides, documents, analysis, and knowledge management. For a worked example of skills chained into one pipeline, see From Academic Paper to Slides.

About

This repository is maintained by Ben Bentzin, Associate Professor at the University of Texas at Austin McCombs School of Business. Ben teaches AI strategy to MBA students and corporate executives and serves as a McCombs Teaching Fellow focused on faculty AI integration. He writes about integrating AI into business at businessai.substack.com.

These skills were built to solve a practical problem: producing consistent, high-quality course materials (slide decks, exercises, case studies, analyses) across a full-semester MBA course on AI strategy. The tools here are the result of that work, generalized for anyone doing similar knowledge work with Claude Code.

What This Is

A collection of Claude Code skills, style guides, and templates that turn ad-hoc AI prompting into structured, repeatable professional workflows. The skills were developed for MBA course development but apply to any knowledge worker who creates documents, presentations, analyses, and research summaries.

Who This Is For

  • Business faculty creating course materials (slides, exercises, cases, discussion outlines)
  • Knowledge workers who produce recurring document types (presentations, proposals, reports)
  • Claude Code users looking for practical skill examples beyond software development
  • Anyone interested in building structured AI environments for professional work

Philosophy

Most AI usage today is ad-hoc: each conversation starts from scratch, corrections are forgotten when the chat ends, and output quality depends entirely on the user's prompting skill in the moment.

A structured AI environment changes this. Corrections persist as feedback memories. Recurring tasks run as reusable skills with built-in quality gates. Project context carries forward through handoffs. The investment is not in learning to code; it is in encoding your professional judgment into rules, skills, and quality standards that compound over time.

This repository publishes the skills, templates, and methodology behind that approach.

The Four-Layer Model

LayerPurposeExample
Rules (CLAUDE.md)Declarative constraints loaded every sessionWriting standards, routing tables, delivery gates
Reinforcement (Memories)Corrections and preferences that persist across sessions"Save each project's deliverables to the top level of its folder"
Execution (Skills)Procedural workflows loaded on demandPDF-to-slides pipeline, exercise generator
Enforcement (Audit Agents)Post-hoc quality checks that catch what the compiler missesBeamer overlay audit, anonymization checklist

See methodology.md for the full design rationale.

Set up your environment

The foundations the rest of this repository builds on. The Getting Started guide walks a newcomer through them, or you can go straight to each:

  • git-sync (recipe). Version your ~/.claude environment with Git and keep two machines in sync through a private GitHub repository. You get history, diffs, and one-command revert on your skills, standards, and memory. Point Claude Code at the directory and it can do the setup for you.
  • handoff-resume (protocol). Carry project context across sessions: end with a structured summary, and start the next session by reading it back.
  • CLAUDE-template. A starting CLAUDE.md to copy and customize for your identity, writing standards, and workflow.

Skills

Task-specific workflows, each loaded on demand when your request matches its triggers. Most install by copying SKILL.md into ~/.claude/skills/<skill-name>/SKILL.md; writing-voice-guide is a process guide you follow to build your own voice layer. Each skill's README documents its own install steps.

SkillDescriptionDetails
split-pdfSplit and deeply read academic PDFs in 4-page chunksAvoids shallow single-pass reads of long documents
beamerGenerate LaTeX Beamer slide decks with TikZ figuresIncludes a four-step compile-audit-fix cycle
slides-contentEnd-to-end: source content to compiled slides to PPTXPDF, markdown, or text input; presentation output
knowledge-baseProcess, index, summarize, and query documentsPersonal knowledge management pipeline
ai-councilFive-advisor council with anonymous peer reviewPressure-test decisions with structured deliberation
ai-council-deepInteractive variant with three user-in-the-loop checkpointsHigh-stakes decisions: term sheets, pivots, pre-publication strategy memos
analyze-replyFact-check a forwarded article, essay, or email and draft a reply in your voicePairs with a writing voice layer
diagram-pdfGenerate standalone TikZ diagrams (pipelines, hierarchies, cycles, hub-and-spoke, thematic)Independent audit agent verifies layout
writing-voice-guideHow to create your own writing voice layerProcess guide, not a skill file

For the full catalog grouped by purpose, including summary-academic, summary-general, knowledge-base-update, and skill-audit, see skills/README.md.

Style Guides

  • Beamer Style Guide: LaTeX setup, color palette, typography, and TikZ conventions
  • PPTX Style Guide: Python setup, PowerPoint formatting rules, and conversion workflow from Beamer

See style-guides/README.md for how to adapt these to your own branding.

Acknowledgments

Several skills in this repository build on work by others:

  • Scott Cunningham, Professor of Economics, Baylor University. split-pdf and the Beamer slide generation approach originated in his MixtapeTools repository. LinkedIn
  • John Graff, Assistant Professor of Instruction, UT Austin McCombs School of Business. The AI Council methodology is adapted from his original LLM Council skill. LinkedIn
  • Andrej Karpathy, Founder, Eureka Labs. The knowledge base pipeline draws on his approach to personal knowledge management. GitHub
  • Freddy Gottesman, GHP Labs. The ai-council-deep interactive variant adds three user-in-the-loop checkpoints (clarify, surface assumptions, iterate) to the AI Council methodology. LinkedIn | GHP Labs

License

MIT

Ver o original no GitHub

Bom para

Relacionados