spring-observability - Claude MCP Skill

# Spring Boot observability

Covers Actuator, Micrometer, the Observation API, tracing, and logging. Details in `references/`. For JDK-level diagnosis (thread dumps, JFR, heap analysis) see the `java-diagnostics` skill; for securing endpoints see `spring-security`.

## Safety rules

- Never read or print secrets (`.env`, credentials, tokens) while inspecting configuration or `/actuator/env`.
- Never expose `heapdump` or `threaddump` over public HTTP. The only exception is a temporary, user-consented window with authentication, TLS and network restriction, removed afterwards (same policy in `java-diagnostics`).
- Never expose `env`, `configprops`, `loggers`, or `shutdown` in production without authentication and network restrictions.
- Pin dependency versions via the Boot BOM; do not add explicit versions for Boot-managed artifacts.

## Detect first

Check the Boot version (3.x or 4.x) and what is already present (`spring-boot-starter-actuator`, `micrometer-registry-*`, tracing/OTel dependencies, `logback-spring.xml`). Boot 4 modularized observability (`spring-boot-micrometer-*`, `spring-boot-opentelemetry`) but starters still bring them in.

## Quick start

```xml
<dependency>
  <groupId>org.springframework.boot</groupId>
  <artifactId>spring-boot-starter-actuator</artifactId>
</dependency>
```

```yaml
management:
  endpoints:
    web:
      exposure:
        include: health,info,prometheus   # allow-list, never "*" in production
  endpoint:
    health:
      probes:
        enabled: true                     # /actuator/health/liveness and /readiness
      show-details: when-authorized
```

## Topics

- [Actuator, health groups and probes, endpoint exposure](references/actuator.md)
- [Micrometer metrics and Prometheus/OTLP export](references/metrics.md)
- [Observation API and tracing with OpenTelemetry](references/tracing-observation.md)
- [Structured logging and correlation ids](references/logging.md)
- [Heapdump/threaddump caveats and runtime diagnosis](references/dumps-and-diagnostics.md)

## Checklist

- [ ] Only needed endpoints exposed and secured; separate management port if possible.
- [ ] Liveness/readiness probes configured for the platform (Kubernetes); readiness group includes critical dependencies only.
- [ ] Metrics exported to the chosen backend; common tags (`application`, `env`) set; no high-cardinality tags (user ids, URLs with ids).
- [ ] Tracing sampling probability set deliberately (Boot default 10%).
- [ ] Logs carry trace/span ids; JSON structured logs in production.
- [ ] Custom observations/metrics have tests.

## Local runtime insights (optional)

For a local-only developer console (and an opt-in MCP server) that shows a running Spring Boot or Quarkus app's runtime state, see BootUI: https://github.com/jdubois/boot-ui and https://www.julien-dubois.com/boot-ui/. Its agent skill can be installed with `gh skill install jdubois/boot-ui skills/bootui`.

- Ask the user for consent before adding any dependency.
- Use it only in a dev/local profile (for example a Maven/Gradle profile or `developmentOnly` configuration); never ship it to production.
- Follow BootUI's own documentation for coordinates and setup rather than copying them here.

See also: <https://docs.spring.io/spring-boot/reference/actuator/observability.html> and <https://docs.micrometer.io/micrometer/reference/>.