convez.ai
SkillProdutividadeIntermediário
jluo41

Tools - Skills Claude Code para Produtividade

jluo41/Tools

Skills Claude Code para produtividade, pesquisa e gestão de conhecimento

Ver no GitHub
InicialPluginScore · de 0 a 100

Coleção de skills personalizados para Claude Code focados em produtividade, pesquisa acadêmica e gerenciamento de conhecimento. Inclui plugins para logging de sessões (Chronicle), diagramas visuais (diagram-skill), pesquisa com IA (HAI-Pipe) e anotação multi-agente (subjective-label).

Instalação

$npx convez-ai jluo41/Tools

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
42%
Último commit
há 2 dias
Versão
—
Issues abertas
0
Contribuidores
5

Casos de uso

  • Logging automático de sessões e rollups diários em Obsidian/JSON Canvas
  • Criação de diagramas ASCII, Excalidraw e draw.io integrados ao fluxo de trabalho
  • Pesquisa acadêmica estruturada com workflows de dados, experimentos e insights
  • Anotação de texto com múltiplos agentes e calibração de critérios subjetivos

Veredito

Conjunto robusto de skills JavaScript bem estruturado para usuários avançados de Claude Code que buscam automação de pesquisa e produtividade. Instalação clara via scripts de marketplace, mas requer familiaridade com Claude Code e gerenciamento local de plugins.

README

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

Tools

Personal Claude Code skill plugins for productivity, academic research, visual artifacts, and knowledge management.

Plugins

PluginDescription
chronicleSession logs, daily rollups, email indexing, and Obsidian/JSON Canvas workflows
diagram-skillASCII diagrams, Excalidraw, draw.io, remote diagram sharing, and progress logs
haipipeHAI-Pipe research toolkit: data, NN, endpoint, task, experiment, insight, paper, and application workflows
subjective-labelMulti-agent subjective text annotation with calibration, validation, and scale workflows

Installation

Quick Setup

git clone git@github.com:jluo41/Tools.git
cd Tools
./install.sh

Then in Claude Code:

/plugin install chronicle@jluo41-tools
/plugin install diagram-skill@jluo41-tools
/plugin install haipipe@jluo41-tools
/plugin install subjective-label@jluo41-tools

By default, install.sh registers this repository as the jluo41-tools marketplace. If the parent workspace has a .claude/ directory, it also symlinks the skills into that workspace.

Windows

Native symlinks on Windows require Administrator or Developer Mode, so use the PowerShell installer, which creates directory junctions instead (no special privileges needed):

cd Tools
.\install.ps1                          # marketplace + auto-detected project skills
.\install.ps1 -Project C:\path\repo    # link into a specific workspace
.\install.ps1 -Global                  # also link into ~\.claude\skills
.\install.ps1 -Symlink                 # use symlinks instead (needs admin / Dev Mode)

Junctions use absolute targets, so re-run install.ps1 if you relocate the repo. The generated links are OS/machine-specific — gitignore <workspace>/.claude/skills/ and regenerate per machine rather than committing them (committed symlinks check out as dead text stubs on Windows).

Project Install

./install.sh --project /path/to/workspace

This symlinks every discovered skill into /path/to/workspace/.claude/skills/. Skill discovery is recursive, so deeply nested skills such as haipipe-toolkit/skills/F_paper/4-write/paper-write are included.

Global Install

./install.sh --global

This symlinks all skills to ~/.claude/skills/ so they are available in every Claude Code session.

Duplicate skill names are resolved deterministically because one skills directory cannot contain two symlinks with the same basename. The installer prefers standalone diagram skills over haipipe utility mirrors and active paper skills over _paper-writing-backup snapshots.

Hooks

./install.sh --hooks

This configures Claude Code sound hooks in ~/.claude/settings.json.

Update

cd Tools
git pull

Skill Development Validation

When adding or changing a skill, treat implementation and validation as two separate steps:

  1. Develop or revise the skill.
  2. Test the skill by calling a subagent with a fresh context and explicitly instructing that subagent to invoke the new or revised skill against a realistic task.
  3. Confirm that the subagent selected and used the skill as expected, followed the skill instructions, and produced the intended result before committing or publishing the change.

This fresh-context subagent check is required because it tests the skill from the point of view of a new agent that has not seen the development discussion. The standard workflow is: develop the skill, then test the skill.

License

MIT

Ver o original no GitHub

Bom para

Relacionados