ODD Team — Manutenção Coordenada por IA
opendatadiscovery/odd-team
Auditoria e manutenção coordenada de múltiplos repos com navegação eficiente.
Ver no GitHubSkill 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-team1 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
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
- Maintain documentation, codebase, and tests in a traceable, maintainable, useful way
- Implement change requests from users — including complexity analysis, dependency mapping, and prioritization
- 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)
| Command | Purpose |
|---|---|
/orient | New 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 |
/status | Progress dashboard |
/navigate <feature> | Find where a feature lives in code |
Target Repositories
Core
| Repo | Purpose |
|---|---|
| odd-platform | Data discovery platform (Java/Spring Boot + React) |
| odd-collectors | Metadata extraction adapters (Python, 40+) |
| opendatadiscovery-specification | ODD data model and Ingress API spec |
| odd-models-package | Generated Python models from spec |
| opendatadiscovery-specification-contracts | Generated code contracts from spec |
Standalone Integrations
| Repo | Purpose |
|---|---|
| odd-great-expectations | Great Expectations plugin — pushes quality results |
| odd-dbt | dbt plugin — pushes model/test/lineage metadata |
| odd-spark-adapter | Spark listener — captures job lineage |
| odd-airflow-2 | Airflow plugin — reports DAG/task metadata |
| odd-collector-profiler | Data profiling collector |
Documentation (Single Source of Truth)
| Repo | Purpose |
|---|---|
| documentation | All user-facing docs — vision, ADRs, features, API, guides |
Deployment, Tooling
| Repo | Purpose |
|---|---|
| charts | Helm charts for Kubernetes deployment |
| odd-cli | CLI for interacting with ODD Platform |
| odd-examples | Usage examples and reference deployments |
| opendatadiscovery-integration-manifests | Registry 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