convez.ai
SkillProdutividadeIntermediário
TC4451

AI Knowledge Base — Agregador de Skills e Aprendizados

TC4451/ai-knowledge-base

Agregue e sincronize skills de IA automaticamente entre projetos

Ver no GitHub
InicialPluginScore · de 0 a 100

Skill que agrega automaticamente conhecimentos e learnings dos diretórios .claude/ da sua máquina e garante que skills de IA estejam sempre instaladas via verificação baseada em manifest. Funciona como base de conhecimento centralizada para Claude Code, permitindo capturar e reutilizar aprendizados sem quebrar o fluxo de trabalho.

Instalação

$npx convez-ai TC4451/ai-knowledge-base

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
2
Forks
0
Acompanhando
0
Saúde
28%
Último commit
há 4 meses
Versão
—
Issues abertas
0
Contribuidores
1

Casos de uso

  • Centralizar conhecimentos e boas práticas de Claude Code entre múltiplos projetos
  • Automatizar instalação de skills via manifest e verificação de dependências
  • Capturar learnings mid-conversation sem interrupção
  • Manter documentação atualizada de conceitos, ferramentas e workflows com IA

Veredito

Ferramenta bem pensada para quem trabalha com Claude Code e precisa gerenciar múltiplas skills e aprendizados. O setup baseado em scripts Shell é robusto e a capacidade de capturar conhecimento mid-conversation agrega valor real ao workflow — embora requeira familiaridade com shell scripting e estrutura de diretórios .claude/

README

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

AI Knowledge Base

📚 A curated collection of AI usage learnings — Claude Code, Kiro, MCP, and more. For both human reference and AI agent consumption.

🔄 Automatically aggregates new knowledge from .claude/ directories across your machine, and ensures your AI skills are always installed via a manifest-based auto-setup check.

TL;DR

# Clone and set up
git clone https://github.com/<your-username>/ai-knowledge-base.git
cd ai-knowledge-base
./scripts/setup.sh          # Install global rule (enables "save this to the KB" from any directory)

# Check skills are installed
diff <(cat skills-manifest.txt) <(ls ~/.claude/skills/)

# Add a learning mid-conversation (works from any directory after setup)
"Save this to the KB — [your learning here]"

# Aggregate knowledge from all .claude/ dirs
./scripts/aggregate.sh

# Regenerate the README after changes
./scripts/generate-readme.sh

Table of Contents

  • Entries — Browse knowledge by category
  • Usage — How to add, aggregate, and manage entries
  • For AI Agents — How agents discover and use this KB
  • Tags — Cross-cutting topic index

Entries

Concepts

  • MCP Tools vs Direct Links Use MCP tools for URLs behind corporate auth; use direct fetch for public URLs.
  • Skills vs Agents in Claude Code Skills are inline slash-commands that run in your conversation; Agents are isolated workers spawned for parallel or noisy tasks.
  • Spec-Driven Development in Kiro Kiro's "spec-based features" means building features through a 3-phase process: Requirements → Design → Tasks, then autonomous implementation.

Tools

Workflows

  • Superpowers Project Workflow The standard project workflow chains skills in order: using-superpowers → brainstorming → planning-with-files → grill-me → writing-plans → executing-plans.

Recipes


Usage

Setup

Run the setup script to register this KB with Claude Code. This creates a global rule at ~/.claude/rules/ so Claude knows where the KB lives — enabling "save this to the KB" from any directory, in any session.

./scripts/setup.sh

After setup, saying "save this to the KB" in any Claude Code session will write directly to this repo, regardless of your current working directory.

Adding Entries

Mid-conversation (easiest):

"Save this to the KB — [describe the learning]"

Claude will create the entry, categorize it, and commit it.

Manually:

  1. Copy templates/entry.md to the appropriate category folder
  2. Rename to a descriptive kebab-case filename (e.g., my-new-learning.md)
  3. Fill in the frontmatter and content sections
  4. Run ./scripts/generate-readme.sh to update the table of contents
  5. Commit your changes

Aggregating from .claude Directories

The aggregator script scans your machine for .claude/ directories and collects memory files, CLAUDE.md content, and rules into the _raw/ inbox.

./scripts/aggregate.sh

By default it searches from $HOME. To search a different root:

./scripts/aggregate.sh /path/to/search

Triaging the inbox: After aggregation, review _raw/ entries manually or ask Claude:

"Triage the _raw/ inbox"

Claude will propose: Categorize (move to proper folder), Skip (delete stale/specific), or Merge (update existing).

Regenerating the README

The table of contents above is auto-generated from entry frontmatter:

./scripts/generate-readme.sh

Run this after adding, moving, or deleting entries.

Categories

FolderWhat goes here
concepts/Mental models, "how X works", clarification of confusing topics
tools/Tool-specific reference (Claude Code, Kiro, MCP servers, etc.)
workflows/Multi-step procedures and processes
recipes/Quick copy-paste patterns and short how-tos

Skill Auto-Setup

This repo tracks which Claude Code skills should be installed globally in skills-manifest.txt. When Claude starts a session in this repo, it checks whether all listed skills are present in ~/.claude/skills/ and reports any that are missing.

To update the manifest:

ls ~/.claude/skills/ > skills-manifest.txt

For AI Agents

Agents discover this KB via:

  • A global rule at ~/.claude/rules/ai-knowledge-base.md (installed by ./scripts/setup.sh) that routes "save this to the KB" commands to this repo from any directory
  • The CLAUDE.md in this repo with detailed instructions when working inside the repo

Agents should:

  1. Read README.md first for the table of contents
  2. Read entries relevant to the current task (use the TL;DR for quick scanning)
  3. Follow CLAUDE.md instructions for saving new entries

Tags

agents ai-coding architecture claude-code customization development-methodology diagnostics enterprise installation kiro knowledge-base mcp note-taking npm obsidian project-lifecycle repo-awareness setup skills specs status-line superpowers terminal urls workflow

Ver o original no GitHub

Bom para

Relacionados