By fppfurtado
Plugin Claude Code que materializa a bridge entre sessões CC e graph Logseq. CLI mb (Python) faz writes mecânicos; 4 skills viram thin orchestrators preservando substância editorial. Captura cross-domain em hashtag-buckets, síntese de sessão com markers GTD, sync repo→Project Page, journal review detective com 4 heurísticas. 2 hooks bridging: Stop sugerindo /journal-close, SessionStart sugerindo /journal-load quando cwd casa com repo owned/active da constelação. Personal-tooling — assume Logseq em ~/Notes/logseq/.
Own this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimOwn this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimBased on adoption, maintenance, documentation, and repository signals. Not a security audit or endorsement.
Cria/atualiza Project Page no graph Logseq a partir do CLAUDE.md/README.md do repo (idempotente; preserva props humanas)
Sintetiza sessão CC no journal Logseq de hoje — narrativa humano-amigável agrupada por
Carrega conteúdo de journals Logseq na sessão CC (read-only) — default journal de hoje; flags --days N retroativo + --bucket
Find-or-create hashtag-bucket no journal Logseq de hoje + append child task com marker GTD opcional
Detective-first review com 4 heurísticas estruturais sobre janela configurável de journals; wizard GTD opt-in (sucessor de /weekly-review v0.2.0)
Claude Code plugin materializing the bridge between CC sessions and a Logseq cognitive graph. Personal-workflow plugin — assumes Logseq at ~/Notes/logseq/ with daily-journal hashtag-bucket convention (- #<domínio> top-level + native GTD task markers) per ADR-006 of meta-system (v0.2.0+).
Companion to pragmatic-dev-toolkit (referenced from the meta-system architecture as Camada 3 — Bridge).
Public for transparency, not for general adoption — most users won't have the assumed setup. If you do (Logseq + hashtag-bucket capture convention), feel free to fork.
A partir de v0.7.0, substância de write vive no CLI mb (Python via Click); 4 das 5 skills viram thin orchestrators que delegam writes ao CLI. Substância heurístico-semântica (matching, princípios editoriais, heurísticas detectivas, cluster prompt) permanece nas skills. Decisão registrada em ADR-002 — materialização por critério target-aware de meta-system ADR-016.
| Component | Type | What it does |
|---|---|---|
/journal-note <content> | Skill (thin orchestrator) → mb journal-note | Skill deriva --domain (git toplevel → basename; senão → AskUserQuestion enum); CLI faz find-or-create top-level - #<domínio> + append child task. Native GTD marker preservado quando input começa com TODO/DOING/WAITING/DONE/CANCELLED . Sub-bullets mecânicos: commit:<hash>, plan:<slug>. Sanitização kebab-case + NFD-strip de acentos PT-BR. Gate pgrep -xi logseq. |
/journal-close | Skill (thin orchestrator) → mb journal-close | Skill faz matching semântico de TODO/WAITING ↔ DONE da sessão CC + síntese editorial humano-amigável agrupada por bucket; CLI é write engine determinístico recebendo payload markdown via stdin (## Append + ## Transitions). Dedup por commit:<hash>, modify-in-place atomic, bootstrap journal. |
/journal-load [--days N] [--bucket #X] | Skill (MD-only) | Read-only — carrega conteúdo de journals na working memory do CC. Permanece markdown-only (sem assimetria CLI > MD; ver ADR-002 § Decisão 5). Default = journal de hoje completo; --days N estende janela retroativa; --bucket #X filtra. Exempt do gate Logseq. |
/init-logseq-project | Skill (thin orchestrator) → mb init-project | CLI faz lookup mecânico de cluster (~/.mrconfig → ~/Projects/meta-system/REPOS.md) + bootstrap via Project Template + props mecânicas idempotentes; skill orquestra AskUserQuestion enum 9-cluster apenas no fallback. Preserva props humanas (status::, description::, blocos ##). |
/journal-review [--days N|--from D --to D] [--interactive] [--write-summary] [--bucket-min-journals N] [--bucket-min-tasks N] [--zombie-days N] [--emerging-min-mentions N] | Skill (thin orchestrator) → mb journal-review | Skill aplica 4 heurísticas semânticas (task-closure-by-context, task-zombie, bucket-underused, bucket-emerging) sobre scan estruturado emitido pelo CLI. Preview-first via AskUserQuestion (Aplicar tudo / Cherry-pick / Cancelar). CLI scan emite markers ativos + DONE-tasks + narrativas + inventário de buckets em ordem cronológica; --apply aplica transições. Skill retém findings em conversation memory entre invocações. Wizard residual via --interactive. 4 flags opcionais de threshold (K=2/M=2/T=21/N=4 defaults) sobrescrevem heurísticas 2b/2c/2d caso-a-caso. Sucessor de /weekly-review. |
suggest_journal_close | Hook (Stop) | Stop event hook, auto-gated triplo (canonical marker [PRAGMATIC: plan-done] emitted by pragmatic-dev-toolkit's /run-plan in its done step + .claude/local/ in cwd + ~/Notes/logseq/ exists AND Logseq desktop closed). When all pass, prints JSON {"systemMessage": ...} to stdout (CC 2.1.x canonical non-blocking soft notification) nudging toward /journal-close. Lógica independente do CLI. |
suggest_session_start_tip | Hook (SessionStart) | SessionStart event hook, gate único cwd-matching: resolve cwd via git rev-parse --show-toplevel, match contra repos owned/active de ~/Projects/meta-system/REPOS.md (filtro NEGATIVO — exclui overview e subsection consumido externo). Match → emit JSON systemMessage sugerindo /journal-load --days 2 --bucket <repo>; no-match/falha → silent. Suite pytest parcial em tests/test_suggest_session_start_tip.py (per ADR-001 Sub-decisão 6 Adendo v0.2.0 + ADR-002 § Decisão 6 Adendo, 2026-06-16). |
mbCLI standalone Python instalável via pipx:
pipx install -e /path/to/meta-bridge
mb --help
npx claudepluginhub fppfurtado/meta-bridge --plugin meta-bridgeClaude Code plugin for the flat & pragmatic development workflow — workflow skills for alignment, execution and diagnosis; YAGNI/QA/security/doc/design reviewers; stack-aware test scaffolder (Python today); self-gated hooks; configurable path contract via CLAUDE.md.
Permanent coding companion for Claude Code — survives any update. MCP-based terminal pet with ASCII art, stats, reactions, and personality.
Intelligent prompt optimization: injects the right context at the right moment so Claude lands a better first output. Clarifies vague prompts with research-based questions, plus targeted nudges for approach selection, plan readability, workflow routing, background execution, subagent routing, output readability, user-decision questions, and plan-mode assessment
Semantic search for Claude Code conversations. Remember past discussions, decisions, and patterns.