AiAi Mastermind — Tools & Skills
philgoodvibe/aiai-mastermind-tools-and-skills
Skills compartilháveis para agentes Claude: wiki-llm e geração de imagens
Ver no GitHubColeção de skills reutilizáveis para agentes como Claude Code, que inclui wiki-llm (constrói uma base de conhecimento Obsidian compilada por IA a partir de documentos) e gpt-image (gera imagens via ChatGPT sem API key). Projetado para ser plugado diretamente em agentes e escalável para múltiplos casos de uso.
Instalação
npx convez-ai philgoodvibe/aiai-mastermind-tools-and-skills1 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
Casos de uso
- Compilar fontes (PDFs, artigos, transcripts) em uma wiki de conhecimento estruturada e interligada
- Gerar imagens de produto, mockups e ilustrações direto do Claude Code via ChatGPT
- Construir bases de conhecimento corporativas, pesquisa acadêmica ou deep-dives temáticos
Veredito

Abordagem sólida de skills modulares com implementação prática de ideias consagradas (Karpathy's llm-wiki). Wiki-llm resolve o problema real de síntese de fontes; gpt-image é criativo mas dependente de OAuth/Codex. Bem documentado e instalável.
README
A tradução para o português deste README ainda está na fila.
AiAi Mastermind — Tools & Skills
A public collection of shareable Agent Skills for Claude Code (and other agents like Codex, Gemini, Aider, Cursor). Drop them into your agent and go.
Skills
🧠 wiki-llm — build a compounding knowledge wiki from your documents
Turn a folder of source documents (books, PDFs, papers, articles, transcripts) into a persistent, interlinked Obsidian knowledge base maintained by an LLM — the opposite of RAG. The LLM reads each source once and compiles it into cross-referenced markdown pages that sit between you and the raw sources, so the synthesis and contradictions are already there and the wiki gets richer with every source you add.
Implements Andrej Karpathy's llm-wiki idea. Domain-agnostic — works for research, a company knowledge base, a single book, a hobby deep-dive, or a business.
It ships with:
- the three-layer RAW → WIKI → OUTPUT architecture and the hard extraction-vs-editorial boundary,
- a domain-agnostic schema template (
CLAUDE.md.template) + five page templates (source / author / concept / framework / synthesis), - ready-to-drop
.obsidian/config — graph color groups, properties, and a CSS snippet that color-codes page types in note view, - the parallel-ingestion fan-out pattern for ingesting dozens of sources at once,
- the raw-traceback protocol (page-cited quotes that trace back to the original file).
🖼️ gpt-image — generate images from Claude Code using your ChatGPT subscription (no API key)
Let Claude Code create images (hero shots, product mockups, illustrations, icons, textures) by delegating to Codex CLI's built-in image_gen tool over your ChatGPT OAuth login — no OPENAI_API_KEY, no per-image API billing. The core insight: a ChatGPT OAuth token is not an API key (the Images API returns 401 missing scope on it), so the skill routes every request through Codex's built-in tool instead of the API.
It enforces the things that are easy to get wrong:
- OAuth, not API — built-in
image_genvia the signed-in ChatGPT session, never a raw API call. - One copy on disk —
mv(nevercp) each output out of Codex's cache, so generated images never silently duplicate. - absolute output paths, one call per image, a network-enabled sandbox, and a fill-in-the-blanks brief template for any use case.
Auto-triggers on "generate / create / make an image"; also ships an optional /gpt-image slash command. Requires the Codex CLI installed and signed in to ChatGPT.
Install
Claude Code (personal skills):
git clone https://github.com/philgoodvibe/aiai-mastermind-tools-and-skills.git
# wiki-llm
cp -R aiai-mastermind-tools-and-skills/skills/wiki-llm ~/.claude/skills/wiki-llm
# gpt-image (+ optional /gpt-image slash command)
cp -R aiai-mastermind-tools-and-skills/skills/gpt-image ~/.claude/skills/gpt-image
cp aiai-mastermind-tools-and-skills/skills/gpt-image/gpt-image.command.md ~/.claude/commands/gpt-image.md
Then just ask: "build a wiki LLM from these PDFs" or "generate a hero image of …" and the matching skill activates. (gpt-image needs the Codex CLI installed and signed in to ChatGPT via codex login.)
Other agents: copy the skill folder into your agent's skills directory (e.g. ~/.agents/skills/ for Codex), or point your agent at skills/wiki-llm/SKILL.md.
License
MIT — see LICENSE. Use it, fork it, share it.
Starter kits
maya-starter/ - Your Automated Social Media Content Team
The complete starter kit for the AIAI Mastermind course "Your Automated Social Media Content Team". It sets up MAYA, the AI agent that runs your content team, with Codex as the copywriter and designer she manages.
Includes: MAYA's standing instructions (CLAUDE.md), a fill-in-the-blanks job description brief, a compliance knowledge-base template, three content-lane skills (newsjacking, problem-solution, entertain), a compliance-check skill, a Mai Marketing Machine posting skill, an image and brand skill, and the brain folder where your My Agency Context document lives.
Setup (covered step by step in the course):
git clone https://github.com/philgoodvibe/aiai-mastermind-tools-and-skills.git
cp -R aiai-mastermind-tools-and-skills/maya-starter ~/Documents/your-agency-content-team
Then open the folder in VS Code, drop your My_Agency_Context.md into brain/, and follow JOB-DESCRIPTION.md.