| name | data-flow-sweep |
| description | Gemini-native end-to-end data flow verification — audits the entire pipeline from producer to dashboard with 2M+ token grounding |
| capability | skill.data_flow_sweep |
| node_name | node_skill_data_flow_sweep_orchestrator |
| dot_color | 208 |
Data Flow Sweep Skill (Gemini Edition)
End-to-end data flow verification for all omnidash projections, leveraging Gemini's massive context window to trace data from its source emission through Redpanda and the consumer into the final dashboard display.
Workflow
- Topic Discovery: Gemini maps all
read_model_topics from topics.yaml to their projection files, DB tables, and dashboard routes.
- Producer Verification: Checks if upstream producers (omniclaude, omnibase_infra, etc.) are actively emitting messages to Redpanda.
- Consumer + DB Verification: Audits consumer group lag and verifies that DB tables are receiving and persisting fresh data.
- Dashboard Page Verification: Optionally uses Playwright to navigate to dashboard routes and ensure data renders correctly without JS errors.
- Remediation: Auto-creates Linear tickets for broken flows, identifying the exact point of failure (Producer, Topic, Consumer, DB, or Page).
Gemini Advantages
- Full-Stack Traceability: Gemini sees the entire pipeline from Python producers to TypeScript/Playwright consumers, identifying bottlenecks that local audits miss.
- Intelligent Lag Analysis: Better at distinguishing between transient lag and structural consumer stalls by analyzing the complexity of the projection logic.
- Cross-Repo Correlation: Correctly identifies when a broken dashboard page is caused by a contract change in a distant producer repository.
Arguments
--dry-run: Report findings without creating tickets.
--topic: Check a single topic only.
--skip-playwright: Skip dashboard page verification.