event-driven-architecture-eda-integrity-evaluator
Evaluate event-driven architectures for payload design, consumer autonomy, broker entanglement, and synchronous callback anti-patterns. Use when reviewing schemas, publishers, consumers, or broker configurations.
Event-Driven Architecture (EDA) Integrity Evaluator
Analyze event-driven systems for autonomy, payload quality, and broker coupling.
When to Use
- PRs that change event schemas, publishers, or consumers
- Message broker configuration changes
- Architectural audits of async or event-driven services
Scope
- Scans: event publishers and consumers, schema registry definitions such as proto/avsc files, and broker configuration for topics, queues, exchanges, or subscriptions.
- Exclusions: direct synchronous APIs, maintenance scripts, and documented notification envelopes that intentionally do not carry full state.
- Fallback: if payload autonomy is unclear, use the shared conventions fallback and report warning with low confidence.
- Routing: defer synchronous workflow concerns to Distributed Workflow and Integration Analyzer and shared-table or read-model ownership to Data and Database Coupling Evaluator.
Workflow
- Analyze payload state.
- Distinguish notification events from event-carried state transfer.
- Flag minimal events that force synchronous lookups after publication.
- Verify consumer autonomy.
- Trace consumer execution for immediate downstream calls.
- Flag consumers that cannot process events from local data alone.
- Check broker entanglement.
- Detect business logic embedded in broker routing or stream configuration.
- Flag overly broad or fat topics with excessive unrelated subscribers.
- Report findings.
- Identify synchronous callback anti-patterns.
- List consumers and topics that violate autonomy.
Rules
- Use static analysis only.
- Inspect schemas, broker config, and source code only.
- Keep analysis deterministic and language-agnostic.
- Prefer bounded-context autonomy over shared middleware logic.
Shared Conventions
- Follow the repository-wide conventions in shared-skill-conventions.
- Use the canonical output contract, severity levels, confidence rubric, routing rules, and false-positive downgrades defined there.
- If this skill is not the closest match, defer to the routing guidance in the shared conventions file.
Output
Return:
- Event topics with synchronous callback risk
- Consumers lacking autonomy
- Broker entanglement findings
- CI-ready pass or fail output