A collection of practical, production-minded workflows that apply Claude to the daily realities of enterprise Java engineering — Spring Boot microservices, Apache Kafka, Oracle, and AWS EKS.
Each workflow is a self-contained module solving a real problem an experienced engineer faces in build, support, and on-call work. They're designed to be genuinely runnable, not toy demos.
| Workflow | Status | What it does |
|---|---|---|
| eks-incident-rca-generator | ✅ Available | Aggregates EKS pod events, logs, HPA history, Kafka lag, and Oracle slow queries, then uses Claude to produce a structured Root Cause Analysis document. |
| oracle-explain-plan-advisor | 🔜 Planned | Interprets Oracle execution plans, flags full table scans / missing indexes, and recommends SQL tuning. |
| kafka-dlq-analyst | 🔜 Planned | Reads a Dead Letter Queue, classifies failure reasons, and suggests remediation + replay strategy. |
| spring-pr-reviewer | 🔜 Planned | GitHub Actions workflow that reviews Spring Boot PRs for transaction, thread-safety, and Kafka consumer pitfalls. |
| kafka-schema-impact-analyzer | 🔜 Planned | Analyzes Avro/Protobuf schema diffs for breaking changes and downstream consumer impact. |
The first workflow is fully built out as the flagship example. The rest are planned modules — this repo is structured as a monorepo so they can be added over time without restructuring.
claude-enterprise-workflows/
├── README.md # you are here
├── CLAUDE.md # guidance for working in this repo with Claude Code
├── LICENSE
├── .github/workflows/ci.yml # CI builds & tests each module
└── eks-incident-rca-generator/ # flagship workflow (Java 21 + Spring Boot)
├── README.md
├── pom.xml
├── src/
├── docs/examples/
└── scripts/
These workflows share a common philosophy:
- Domain context over generic prompting. Claude is given senior-engineer system prompts with real Spring/Kafka/Oracle/EKS failure patterns, so its reasoning reflects the stack — not generic advice.
- Works in locked-down enterprises. The LLM backend is pluggable (internal gateway/proxy, direct API, or fully offline), and signal collection is decoupled from analysis so a host with infrastructure access never needs outbound internet.
- Best-effort signal collection. A missing data source degrades gracefully rather than failing the whole run.
- Structured output contracts. Claude is asked for strict JSON, parsed defensively so a slightly off response still yields a usable result.
- Human-in-the-loop. Every output is a draft for an engineer to review — these tools accelerate judgment, they don't replace it.
- Secrets stay in the environment. No API keys, gateway tokens, or DB passwords in config files.
- Java 21, Spring Boot 3.2
- Anthropic Java SDK
- Fabric8 Kubernetes client, Kafka AdminClient, Oracle JDBC
- picocli for CLI, FreeMarker for templating
Each module has its own README with build and run instructions. Start with the flagship:
cd eks-incident-rca-generator
mvn clean package
cat README.md- Create a new top-level module directory.
- Add it to the workflow table above.
- Add a build job to
.github/workflows/ci.yml. - Include a module
README.mdand adocs/examples/folder with sample input/output.
See LICENSE.