convez.ai
SkillDevOps & InfraestruturaAvançado
opendatadiscovery

ODD Team — Manutenção Coordenada por IA

opendatadiscovery/odd-team

Auditoria e manutenção coordenada de múltiplos repos com navegação eficiente.

Ver no GitHub
InicialPluginScore · de 0 a 100

Skill que coordena um time de maintainers IA para o projeto Open Data Discovery, automatizando auditoria de documentação, testes e cobertura de código. Fornece scanners rastreáveis, índices de navegação e gestão de backlog estruturada para manter repositórios grandes em estado sustentável.

Instalação

$npx convez-ai opendatadiscovery/odd-team

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
50%
Último commit
há 2 dias
Versão
—
Issues abertas
0
Contribuidores
2

Casos de uso

  • Mapear gaps em documentação, testes e cobertura de código automaticamente
  • Coordenar implementação de mudanças com análise de dependências e priorização
  • Manter índices de navegação para encontrar features em O(1) sem varreduras inteiras

Veredito

Solução robusta para governança de grandes projetos open-source com múltiplos repos. Exige disciplina na estrutura de scanners e índices, mas libera tempo significativo em auditoria e rastreamento de débito técnico. Produção-ready para equipes que tratam manutenção como atividade sistematizada.

README

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

ODD Team

A team of AI maintainers for the Open Data Discovery project.

This repo is an agent workspace — scanners, work items, navigation pointers, progress tracking. It is NOT user-facing documentation. The single source of truth for all docs is opendatadiscovery/documentation.

Mission

  1. Maintain documentation, codebase, and tests in a traceable, maintainable, useful way
  2. Implement change requests from users — including complexity analysis, dependency mapping, and prioritization
  3. Answer questions about features, implementations, and deployment options

Current Phase: Audit & Gap-Closing

Before the team can maintain and extend the project, it needs a solid foundation: accurate docs, adequate test coverage, navigable code, and correct specifications. The current focus is systematically closing these gaps across all ODD repositories.

Workflow

/scan <scanner>       → discover gaps, write findings, enrich navigation index
/triage <findings>    → convert to reviewable work items
[human reviews]       → approve/reprioritize backlog
/implement <item-id>  → execute change, atomic commit in target repo
/review <item-id>     → verify acceptance criteria

Architecture

navigation/        Feature-to-code pointer maps (token-efficient, prevents redundant scanning)
scanners/          Audit definitions (docs, tests, navigation, spec)
findings/          Raw audit output (timestamped per scanner)
backlog/           Triaged work items (DOC/TST/NAV/SPC)
state/             Progress dashboard, file registry, decisions
agents/            Reusable role prompts
.claude/skills/    Slash commands for each workflow step

Navigation Index

The codebase is large — multiple repos, hundreds of files per feature. The navigation index (navigation/) provides token-efficient pointer maps so any AI session can find relevant code in O(1) reads without scanning entire repositories. This is the foundation that makes all other capabilities sustainable.

Skills (Slash Commands)

CommandPurpose
/orientNew session briefing — state and recommended next action
/enumerate <scanner>List ALL items in a scanner's scope (generates coverage manifest)
/scan <scanner>Scan next unscanned batch from coverage manifest
/coverage [scanner]Show what's been analyzed vs. what remains
/triage <findings>Convert findings to work items
/implement <id>Execute an approved work item
/review <id>Verify completed work
/statusProgress dashboard
/navigate <feature>Find where a feature lives in code

Target Repositories

Core

RepoPurpose
odd-platformData discovery platform (Java/Spring Boot + React)
odd-collectorsMetadata extraction adapters (Python, 40+)
opendatadiscovery-specificationODD data model and Ingress API spec
odd-models-packageGenerated Python models from spec
opendatadiscovery-specification-contractsGenerated code contracts from spec

Standalone Integrations

RepoPurpose
odd-great-expectationsGreat Expectations plugin — pushes quality results
odd-dbtdbt plugin — pushes model/test/lineage metadata
odd-spark-adapterSpark listener — captures job lineage
odd-airflow-2Airflow plugin — reports DAG/task metadata
odd-collector-profilerData profiling collector

Documentation (Single Source of Truth)

RepoPurpose
documentationAll user-facing docs — vision, ADRs, features, API, guides

Deployment, Tooling

RepoPurpose
chartsHelm charts for Kubernetes deployment
odd-cliCLI for interacting with ODD Platform
odd-examplesUsage examples and reference deployments
opendatadiscovery-integration-manifestsRegistry of supported integrations

Roadmap

Phase 1: Foundation (current)

  • Audit documentation accuracy, completeness, coverage
  • Audit test coverage across platform and collectors
  • Build navigation index for all feature domains
  • Verify specification alignment
  • Close all identified gaps

Phase 2: Change Request Implementation

  • Accept change requests as GitHub Issues
  • Perform complexity and dependency analysis
  • Break down into implementable tasks with prioritization
  • Execute with full traceability (Issue → PR → commit)

Phase 3: Knowledge & Support

  • Answer user questions about features and architecture
  • Provide deployment guidance
  • Explain implementation decisions with references to code and ADRs
  • Community support via Slack integration

Principles

  • Backward-incompatible changes avoided unless absolutely necessary
  • Chase engineering perfection honestly — log imperfections as issues, fix incrementally
  • Heavily rely on PostgreSQL-native capabilities
  • No functional changes during audit phase — only docs, tests, comments, spec alignment
  • Human reviews backlog before implementation
  • Navigation index prevents redundant scanning across sessions
  • One work item = one atomic commit
Ver o original no GitHub

Bom para

Relacionados