convez.ai
SkillProdutividadeAvançado
Boming0002

Work Memory Vault

Boming0002/work-memory-vault

Claude como bibliotecário de seu Obsidian com organização PARA automática e auditoria

Ver no GitHub
InicialPluginScore · de 0 a 100

Framework de skills testado em produção que transforma Claude em bibliotecário de seu cofre Obsidian. Ingere qualquer conteúdo (transcrições, tarefas, mensagens) e o organiza automaticamente em estrutura PARA com frontmatter consistente, wiki-links e dashboard auditável. Inclui 5 skills principais, módulos compartilhados, scripts de manutenção e proteção contra escrita malformada.

Instalação

$npx convez-ai Boming0002/work-memory-vault

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
0
Forks
0
Acompanhando
0
Saúde
42%
Último commit
há 5 meses
Versão
—
Issues abertas
0
Contribuidores
1

Casos de uso

  • Organizar automaticamente transcrições de reuniões em notas estruturadas com decisões e ações
  • Rastrear e gerenciar tarefas datadas e atribuídas com revisão semanal de higiene
  • Manter um cofre de conhecimento sem degradação com varreduras de drift e detecção de arquivos obsoletos
  • Implementar gate de revisão que exige aprovação humana antes de qualquer escrita no vault

Veredito

Solução robusta e testada em produção para gerenciar conhecimento pessoal com Claude, mas requer familiaridade com Obsidian, linha de comando e arquitetura de skills. Excelente para quem precisa de um sistema confiável de ingestão e organização com garantias de integridade.

README

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

work-memory-vault

A battle-tested skill framework that turns Claude Code into the librarian of your Obsidian work vault.

Paste anything — a meeting transcript, a task from your boss, a teammate's chat dump — and the right skill files it into a PARA-structured vault with consistent frontmatter, wiki-links, person stubs, and a dashboard that stays honest. Every write goes through a review gate. A weekly hygiene pass and drift scanners keep the vault from rotting.

This is not a starter template: it's the extracted core of a vault that runs in daily production use, including the failure modes (the F-code catalog) it learned the hard way and the architecture decisions (docs/adr/) behind every non-obvious choice.

What you get

PieceWhat it does
5 skills/intake (file anything) · /meeting (transcripts → notes + decisions + actions) · /task (tracked, dated, owned todos) · /weekly-review (hygiene pass: inbox, stale tasks, drift scans) · /daily-note (daily scaffold + end-of-day wrap)
6 shared modulesreview-gate confirmation rules, parallel post-write audits, raw-source disposition, pinned-dashboard curation, path classification, and a catalog of real failure modes (F-codes)
2 maintenance scriptsvault-drift-scan.sh (4 read-only drift checks incl. cold-file detection) · skill-usage-week.sh (usage metrics + honesty checks on the skills themselves)
Write-guard hookoptional PreToolUse hook that blocks malformed vault writes (missing frontmatter, undated tasks, forbidden zones)
Vault templatePARA skeleton + Templater templates + dashboard home page + a CLAUDE.md vault constitution you adapt
ADRs12 published architecture decisions — why symlinks, why a review gate, why the skills never decide for you

Quickstart (5 minutes)

Prereqs: Claude Code, Obsidian (optional but intended), macOS/Linux.

git clone https://github.com/Boming0002/work-memory-vault.git
cd work-memory-vault
./install.sh          # prompts for your vault path (or scaffolds a new one)

Then in any Claude Code session:

/intake    ← paste any work content, watch it get filed properly

Full walkthrough (Obsidian plugins, dashboard, cloud sync, the hook): docs/SETUP.md How the three layers fit together: docs/ARCHITECTURE.md

Design principles (the short version)

  1. The vault is yours; Claude is the librarian. Skills propose, you approve — one y per logical batch, never silent writes.
  2. Subagents collect data, never decide. Post-write audits run in parallel; anything requiring judgment comes back to you.
  3. Hygiene is mechanized, not hoped for. Weekly review + drift scans + a write-guard hook catch rot before it spreads.
  4. Failures become F-codes. Every real mistake the skills ever made is cataloged with its avoidance rule — the skills cite them inline.

License

MIT © Boming Jiang

Ver o original no GitHub

Bom para

Relacionados