| name | pipeline-audit |
| description | Gemini-native end-to-end multi-repo pipeline audit — system-wide integration verification with 2M+ token grounding |
| capability | skill.pipeline_audit |
| node_name | node_skill_pipeline_audit_orchestrator |
| dot_color | 208 |
Pipeline Audit Skill (Gemini Edition)
Systematically audit an end-to-end multi-repo pipeline for integration correctness, leveraging Gemini's massive context window to prove every join between services with file-level evidence.
Workflow
- Repository Discovery: Gemini identifies every repository participating in the pipeline and analyzes its purpose.
- Capability Inventory: Dispatches parallel Gemini-native agents to build structured maps of integration surfaces (Kafka, DB, API, Config) per repo.
- Pipeline Trace: Gemini constructs an end-to-end pipeline trace, mapping narrative flows to exact wire formats.
- Hard Assertions: Proves integration correctness across six categories:
- Runtime Entrypoint: Verification that services actually start and reach integration code.
- DSN Proof: Confirmation that all repos connect to the same database/schema.
- Wire Topics: Byte-for-byte matching of producer and consumer topic strings.
- Schema Handshake: Agreement on table schemas across writers, readers, and dashboards.
- Wire Format Compatibility: Compatibility verification of message serialization/deserialization.
- Correlation ID Threading: Tracing of identifier propagation through every stage.
- Gap Register: Compiles all findings into a severity-ordered list of integration issues.
- Ticket Creation: Auto-creates Linear tickets for identified gaps with exact code references and remediation guidance.
Gemini Advantages
- Whole-Pipeline Grounding: Gemini can simultaneously verify that a producer in repo A matches a consumer in repo B across Kafka, DB, and API boundaries.
- Superior Pattern Context: Better at identifying "broken links" in the correlation chain by understanding the technical intent of the tracing identifiers.
- Context-Aware Schema Audit: More accurately identifies breaking changes in message models by seeing the entire organization's Pydantic config.
Arguments
PIPELINE_NAME: Name of the pipeline to audit.--all: Run all 6 proof categories.--dry-run: Preview findings without creating tickets.
