Every skill and every agent
A generated project ships 32 skills and 7 agents. Each has a page of its own: what it reads and writes, what the guard does with it, and where it sits on the chains. This page is the way in.
Skills
19 skills in two groups: the design skills of /new-feature, and the pieces of /init-project with /git-publish.
design · /new-feature 13
use-case-designTurns a request into one use case: its boundary, flow and canonical names.
domain-modelingModels the aggregate, its invariants, ports and events from the use case.
rest-api-architectDesigns the HTTP contract: endpoints, DTOs, error map, pagination, idempotency and versioning.
graphql-api-architectDesigns the GraphQL adapter: schema, inputs and payloads, connections, batch loads and the error map.
security-architectDesigns who may call what: the mechanism, authorities, the filter chain and the failure responses.
http-client-architectDesigns the outbound HTTP adapter: client, pool, timeouts, retry, resilience and outbound authentication.
ai-architectDesigns the AI ports: the model or vector store, prompts, the answer type, data sent and failure families.
search-architectDesigns the search index as a rebuildable projection: mapping, queries, writes, rebuild and lag.
messaging-architectDesigns the Kafka adapters: topic, partition key, delivery, consumer idempotency, retry and DLQ.
jobs-architectDesigns the scheduled and background jobs: technology, coordination across instances and execution guarantees.
persistence-architectDesigns how the aggregate reaches disk and comes back: schema, mapping, adapters and migrations.
cache-architectDesigns the application cache as a decorator of a port: keys, provider, expiry, invalidation and failure.
test-architectDesigns the tests: what is tested at which level, the data, the doubles and the coverage gaps.
/init-project and /git-publish 6
git-publishOffers to commit and push the project, behind two explicit confirmations.
init-projectRuns the interview of a project's first generation in the main session, then hands the whole generation to one agent.
project-bootstrapTurns a new directory next to nerviz into a Spring Boot project from a blueprint: Initializr base, boundaries, hooks and a green build.
transport-security-setupDecides where TLS terminates, at the edge or in the embedded Tomcat, and writes the matching configuration and the test that proves it.
sonarqube-setupWires the project to a SonarQube server: the scanner in the build file, and a CI step or a local container.
docker-architectOwns docker-compose.yml and the Dockerfile: adds the services a project or a use case needs.
Commands and tools
13 skills that sit on neither chain.
new-featureOrchestrates the design of a feature: one use case per run, the design skills in order, to an approved spec the executor can build.
new-feature-implementImplements one approved spec: a setup pre-flight, then the executor in three chained block groups, then it prints the git-publish command.
arch-adoptInstalls or updates this architecture's .claude/ in a Java project that /init-project did not generate: norms, skills, agents, the hook and its schema.
platform-upgradeTakes a Maven or Gradle project on a Spring Boot line below the floor up to it, and leaves the build green and the diff uncommitted.
new-projectPrepares a new project's generation without a clone: fetches the source at its latest tag, records the ref and commit, and hands over to /init-project.
project-configShows and changes what a project configures in its own .claude/, through ArchHook.jar config, the only writer.
arch-doctorDiagnoses the setup on this machine: hooks, boundary rules, the build wrapper, java, which .claude/ the project came from, and whether the compose stack runs.
audit-usageReads the execution trail across runs: billable tokens per skill and agent, duration, failures, permissions, and which report to open next.
report-issueFiles an issue on the Nerviz repository from inside a generated project, from a description or a lessons-learned file, and refuses a report with no evidence a maintainer can verify.
triage-issueTriages one issue skeptically: every claim goes to the read-only issue-verifier, and a verdict is posted only after you confirm.
sonar-lessonsRuns the project's SonarQube analysis, traces every group of findings to the norm or setting that produced it, and writes a lessons file.
gof-design-patternsChooses and implements a design pattern in Java and Spring from the symptom that justifies it.
claude-code-architect-designerDecides which Claude Code extension resolves a scenario (skill, subagent, rule, hook, MCP server, permission or mod) and writes the files only after you approve.
Agents
7 subagents. A skill hands work to one, which runs in a fresh context with its own tools and model. Its writes are judged by its own class.
commons-logging-installerInstalls the logging and masking annotations and aspects into the project's empty commons package, once.
mapstruct-installerRecords the project's mapper mechanism once: writes the manual-mappers paragraph, or wires MapStruct and converts every existing mapper.
java-spring-boot-developerImplements an approved spec.
compose-service-installerAdds to docker-compose.yml every service the spec marks materialize, by running docker-architect.
project-initializerDrives a project's first generation from the interview's answers: follows project-bootstrap, installs the hooks and verifies the build. Asks nothing.
archunit-installerInstalls ArchUnit and wires the JaCoCo coverage gate into an already-generated project. Invoked by test-architect's setup mode, never by you.
issue-verifierVerifies every claim of one issue against the code at HEAD, without taking its diagnosis or fix as true, and returns a verdict claim by claim. Invoked by /triage-issue, never by you.