geneseed 4.9.4 → 4.10.0

package/js/web/actions.mjs

@@ -34,7 +34,7 @@ import path from 'node:path';
 
 import { emitGlobalInto } from '../build/driver.mjs';
 import {
-  ROOT, discoverNames, knownRuleIds, PACK_ORDER,
+  ROOT, discoverNames, knownRuleIds, loadKits, PACK_ORDER,
 } from '../build/source.mjs';
 import { withStdoutSwallowed } from '../inspect/diff.mjs';
 import { excludeAdd, excludeRemove } from '../inspect/excludes.mjs';
@@ -46,9 +46,10 @@ import {
   opencodeConfigDir, resolvePath,
 } from '../hosts/hosts.mjs';
 import {
-  EMIT_HOST_SCOPE, doctrinesForBuild, excludedRulesOfDir, excludedSkillsOfDir,
+  EMIT_HOST_SCOPE, doctrinesForBuild, excludedRulesOfDir, excludedSkillsOfDir, kitsOfDir,
   footprintOfDir, gatesOfDir,
-  installState, installTargets, legacyProcessCarrierOfDir, modeOfDir, postureOfDir, trustOfDir,
+  installState, installTargets, legacyProcessCarrierOfDir, modeOfDir, postureOfDir, themeOfDir,
+  trustOfDir,
 } from '../hosts/installs.mjs';
 import {
   isDict, mcpCommented, mcpInstallTargets, mcpLoad, mcpServerNames, mcpState,
@@ -131,6 +132,15 @@ function bodyExcludeRules(body) {
     (n) => n.trim().replace(/[ \t]+/, '.'));
 }
 
+/**
+ * A request body's `kits` — REPLACES the install's kits (the Kits tab sends the whole set it
+ * wants): `[]`/`'none'` give `[]`, an omitted key gives `null` (keep the install's own), and an
+ * unknown name gives `null` too, which `apiInstallCmd` refuses for a body that HAS the key.
+ */
+function bodyKits(body) {
+  return bodyList(body, 'kits', () => [...loadKits().keys()]);
+}
+
 /**
  * The gate switches across the trust boundary (`deletionGate`, `secretsGate`, `doneGate`), each
  * on its own: only a real JSON boolean counts. Anything else — a string `"false"`, a number,
@@ -539,6 +549,8 @@ export function apiRestore(state, files) {
       // re-states a rule whose hook gate the install deliberately left unwired.
       excludeRules: excludedRulesOfDir(target),
       excludeSkills: excludedSkillsOfDir(target),
+      // The kits: a restored AGENT.md keeps its Kits block and `Active kits:` line.
+      kits: kitsOfDir(target),
       // The gate switches: a restored AGENT.md must keep the install's `<gate>: off` lines.
       ...gatesOfDir(target),
     }));
@@ -607,8 +619,9 @@ export function apiInstallCmd(state, body) {
   if (emit === undefined) return { error: `no install mode for ${host}:${scope}` };
   // Every read-back below is narrowed to `host`'s OWN carrier (host-compat B1) — see the
   // comment at `installProfile` in `js/build/generate.mjs`.
+  // The theme default is THIS install's own, not the viewed one's: the Kits tab acts on any row.
   const { theme, footprint: fp, posture: pos, mode, trust } = bodyAxes(body, {
-    theme: state.theme, footprint: footprintOfDir(root),
+    theme: themeOfDir(root, host) || state.theme, footprint: footprintOfDir(root),
     posture: postureOfDir(root, host) || 'peer',
     mode: modeOfDir(root, host) || 'direct', trust: trustOfDir(root, host) || DEFAULT_PRESET });
   // Unspecified means "keep what this install already has", exactly as theme, footprint,
@@ -626,9 +639,20 @@ export function apiInstallCmd(state, body) {
   // The gate switches (the Constitution page's toggles): a strict boolean per gate in the body,
   // else the install's own answer — so a pack Apply keeps every switch as it is.
   const gates = bodyGates(body, root, host);
+  // The page sends `kits` deliberately, so an unknown name is refused: ignoring it would run a
+  // no-op rebuild and report success. An omitted key keeps the install's own kits.
+  let kits = null;
+  try {
+    kits = bodyKits(body);
+    if (kits === null && bget(body, 'kits') !== null) {
+      return { error: `unknown kit in ${formatRepr(bget(body, 'kits'))} — this Geneseed ships: `
+        + [...loadKits().keys()].join(', ') };
+    }
+  } catch (e) { return { error: String(e?.message ?? e) }; }
   const out = scope === 'global' ? null : String(root);
   const argv = setupBuildArgs(theme || 'neutral', emit, out, out, fp, pos, mode, doctrines,
-    PACK_ORDER, excludeRules, trust, excludedSkillsOfDir(root, host), gates);
+    PACK_ORDER, excludeRules, trust, excludedSkillsOfDir(root, host), gates,
+    kits ?? kitsOfDir(root, host));
   return { cmd: [process.execPath, path.join(ROOT, 'bin', 'build-driver.mjs'), ...argv] };
 }
 
@@ -730,6 +754,6 @@ export function apiDeployCmd(state, body) {
   const gates = bodyGates(body, root, host);
   // project-scope emit name == host name (opencode / claude / bob / openclaude)
   const argv = setupBuildArgs(theme || 'neutral', host, root, root, fp, pos, mode, doctrines,
-    PACK_ORDER, excludeRules, trust, excludedSkillsOfDir(root, host), gates);
+    PACK_ORDER, excludeRules, trust, excludedSkillsOfDir(root, host), gates, kitsOfDir(root, host));
   return { cmd: [process.execPath, path.join(ROOT, 'bin', 'build-driver.mjs'), ...argv] };
 }

package/js/web/api.mjs

@@ -11,22 +11,24 @@
 import { statSync } from 'node:fs';
 import path from 'node:path';
 
-import { CONFIG, ROOT, THEMES, discoverNames, makeCfg } from '../build/source.mjs';
+import { CONFIG, ROOT, THEMES, discoverNames, loadKits, makeCfg } from '../build/source.mjs';
 import { readVersion, sourceFingerprint } from '../build/version.mjs';
 import { sourceReleaseVersion } from '../hosts/opencode.mjs';
 import { alwaysOnCost } from '../inspect/cost.mjs';
+import { installProfile } from '../build/generate.mjs';
 import { diffCollect, renderMemo } from '../inspect/diff.mjs';
 import { excludesSnapshot } from '../inspect/excludes.mjs';
 import { CLAUDE_STYLE, HOSTS, VERSION_MARKER, resolvePath } from '../hosts/hosts.mjs';
 import {
-  footprintOfDir, gatesOfDir, installState, installTargets, modeOfDir, trustOfDir,
+  footprintOfDir, gatesOfDir, installState, installTargets, legacyProcessCarrierOfDir, modeOfDir,
+  trustOfDir,
   postureOfDir, readJsonMaybe, themeOfDir,
 } from '../hosts/installs.mjs';
 import { PRESETS } from '../loop/score.mjs';
 import { loadCatalog, catalogProblems } from '../loop/catalog.mjs';
 import { activeLoops } from '../loop/registry.mjs';
 import { EMIT_OPTIONS, themeOptions } from '../maintain/setup.mjs';
-import { accentFor, statusData } from '../inspect/status.mjs';
+import { accentFor, keptKitFiles, statusData } from '../inspect/status.mjs';
 import { isFile } from '../lib/fs.mjs';
 import { comparePaths, normcase } from '../lib/paths.mjs';
 import {
@@ -231,6 +233,41 @@ export function apiFleet(state) {
   return { installs, source_fp: sourceFp, doctor: state.doctor };
 }
 
+/**
+ * EVERY INSTALL'S KITS — the Kits tab. Its own route, like `apiFleet`: it re-reads each carrier
+ * and the kit manifests, which the sidebar's `/api/installs` refetch must not pay for.
+ *
+ * `kits` per row is `installProfile(...).kits`, the reader `rebuild-all` uses, so the tab shows
+ * what a rebuild would keep. A malformed manifest answers 200 with `error` and EMPTY lists, never
+ * rows without kits (which would read as "no kit on") and never a 500. `editable:false` carries
+ * the two refusals `apiInstallCmd` makes, so the page can say why before the click.
+ */
+export function apiKits() {
+  let known;
+  try { known = loadKits(); } catch (e) {
+    return { kits: [], installs: [], error: String(e?.message ?? e) };
+  }
+  const kits = [...known].map(([name, m]) => ({
+    name, description: m.description ?? '', skills: m.skills ?? [], agents: m.agents ?? [] }));
+  const installs = [];
+  for (const [host, scope, root] of installTargets()) {
+    const st = installState(root, host, scope);
+    if (st === 'absent') continue;
+    try {
+      const p = installProfile(host, scope, root);
+      const reason = st === 'disabled' ? 'install is disabled — reactivate it first'
+        : legacyProcessCarrierOfDir(root, host) ? 'predates the comms pack — run geneseed setup for it first'
+          : null;
+      installs.push({ ...installRowBase(host, scope, root, st), kits: p.kits,
+        keptKitFiles: keptKitFiles(host, scope, root, p.kits), editable: reason === null, reason });
+    } catch (e) {
+      installs.push({ ...installRowBase(host, scope, root, 'unreadable'),
+        error: String(e?.message ?? e) });
+    }
+  }
+  return { kits, installs };
+}
+
 /** The dashboard aggregate. */
 export function apiOverview(state) {
   const inv = state.inventory;

package/js/web/routes.mjs

@@ -21,7 +21,7 @@ import { percentDecode } from '../lib/text.mjs';
 import { apiExcludesMutate, apiInstallToggle, apiLoopsPresetMutate, apiMcp, apiMemoryDelete, apiSelectView } from './actions.mjs';
 import { apiActivity, apiActivityDetail, apiActivityToggle } from './activity.mjs';
 import {
-  apiDiff, apiDoctor, apiExcludes, apiFleet, apiInstalls, apiLoops, apiLoopsActive, apiOverview,
+  apiDiff, apiDoctor, apiExcludes, apiFleet, apiInstalls, apiKits, apiLoops, apiLoopsActive, apiOverview,
   apiRecent, apiSetup, apiThemes, apiUpdate,
 } from './api.mjs';
 import { NotFound, apiCatalog, apiItem } from './catalog.mjs';
@@ -151,6 +151,7 @@ export const STATE_ROUTES = {
     new URL(url || '', 'http://x').searchParams.has('fresh')),
   '/api/installs': apiInstalls,
   '/api/fleet': apiFleet,
+  '/api/kits': apiKits,
   '/api/update': apiUpdate,
   '/api/usage': apiUsage,
   '/api/gates': apiGates,

package/package.json

@@ -1,7 +1,7 @@
 {
   "//": "Every field here is argued and gated in tests/unit/package_manifest.test.mjs. `files`, `bin` and `engines` each have a test that fails when reality drifts from this file; `version` mirrors harness.config.json and is gated too. Read that docblock before editing this one.",
   "name": "geneseed",
-  "version": "4.9.4",
+  "version": "4.10.0",
   "description": "A portable, theme-able harness you implant once and use everywhere to grow a disciplined AI coding agent.",
   "keywords": [
     "ai",

package/registry.json

@@ -22,6 +22,20 @@
       "added": "2026-07-13",
       "last_verified": ""
     },
+    "agents/devops-engineer": {
+      "status": "experimental",
+      "version": "1.0.0",
+      "owner": "Guillaume Lemer",
+      "added": "2026-10-20",
+      "last_verified": ""
+    },
+    "agents/docker-expert": {
+      "status": "experimental",
+      "version": "1.0.0",
+      "owner": "Guillaume Lemer",
+      "added": "2026-10-20",
+      "last_verified": ""
+    },
     "agents/docs": {
       "status": "approved",
       "version": "1.0.0",
@@ -57,6 +71,13 @@
       "added": "2026-06-08",
       "last_verified": ""
     },
+    "agents/kubernetes-specialist": {
+      "status": "experimental",
+      "version": "1.0.0",
+      "owner": "Guillaume Lemer",
+      "added": "2026-10-20",
+      "last_verified": ""
+    },
     "agents/memory-curator": {
       "status": "experimental",
       "version": "1.0.0",
@@ -113,6 +134,27 @@
       "added": "2026-06-08",
       "last_verified": ""
     },
+    "agents/spring-boot-engineer": {
+      "status": "experimental",
+      "version": "1.0.0",
+      "owner": "Guillaume Lemer",
+      "added": "2026-10-20",
+      "last_verified": ""
+    },
+    "agents/spring-reviewer": {
+      "status": "experimental",
+      "version": "1.0.0",
+      "owner": "Guillaume Lemer",
+      "added": "2026-10-20",
+      "last_verified": ""
+    },
+    "agents/spring-security": {
+      "status": "experimental",
+      "version": "1.0.0",
+      "owner": "Guillaume Lemer",
+      "added": "2026-10-20",
+      "last_verified": ""
+    },
     "agents/steward": {
       "status": "approved",
       "version": "1.0.0",
@@ -428,6 +470,27 @@
       "added": "2026-09-06",
       "last_verified": "2026-10-13"
     },
+    "skills/spring-boot": {
+      "status": "experimental",
+      "version": "1.0.0",
+      "owner": "Guillaume Lemer",
+      "added": "2026-10-20",
+      "last_verified": ""
+    },
+    "skills/spring-data-jpa": {
+      "status": "experimental",
+      "version": "1.0.0",
+      "owner": "Guillaume Lemer",
+      "added": "2026-10-20",
+      "last_verified": ""
+    },
+    "skills/spring-testing": {
+      "status": "experimental",
+      "version": "1.0.0",
+      "owner": "Guillaume Lemer",
+      "added": "2026-10-20",
+      "last_verified": ""
+    },
     "skills/teach": {
       "status": "experimental",
       "version": "1.0.0",

package/src/AGENT.md.tmpl

@@ -145,7 +145,7 @@ Inactive packs still ship under `{{DIR_DOCTRINES}}/`, so a citation into one res
 {{DOCTRINES_BODY}}
 
 Active packs: {{DOCTRINES_LIST}}{{EXCLUDED_RULES_LINE}}{{GATE_LINES}}
-
+{{KITS_BODY}}
 ---
 
 ## 3. {{AGENTS}} — delegation by capability
@@ -288,7 +288,7 @@ inventory — that list is the catalogue and it is authoritative. Each written
 procedure is in [`{{DIR_SKILLS}}/`]({{DIR_SKILLS}}/) and the host loads it on
 demand.
 <!-- CATALOG:end -->
-{{EXCLUDED_SKILLS_LINE}}
+{{EXCLUDED_SKILLS_LINE}}{{KITS_LINE}}
 When a task repeats and no {{SKILL}} covers it, forge one through the
 [skill-forge]({{DIR_SKILLS}}/skill-forge.md) {{SKILL}}: copy
 [`{{DIR_SKILLS}}/_template.md`]({{DIR_SKILLS}}/_template.md), keep it to one coherent

package/src/agents/devops-engineer.md

@@ -0,0 +1,45 @@
+# {{AGENT}}: devops-engineer
+
+> {{DESC_DEVOPS_ENGINEER}}
+
+{{CREDIT:piomin-spring-boot:agents/devops-engineer.md}}
+
+Builds CI/CD pipelines and delivery automation for Spring Boot applications.
+
+## When to dispatch
+- A Spring Boot project needs a pipeline created or fixed: build, test, image, scan, deploy stages.
+
+## When NOT to dispatch
+- Image internals — use [docker-expert](docker-expert.md). Cluster manifests — use [kubernetes-specialist](kubernetes-specialist.md).
+
+## Inputs
+- The repository, its existing CI provider and the target registry and environments.
+
+## Allowed tools
+- **Read + write to the assigned worktree.** May run the build and pipeline linters.
+  <!-- bash: allow -->
+- Never commits, pushes or triggers a deployment ({{DOCTRINE:consent-before-push}}).
+
+## Procedure
+0. If `{{DIR_MEMORY}}/agents/<your-name>.md` exists, read it first — your durable lessons from prior dispatches ({{DOCTRINE:persist-insight}}).
+1. Detect the CI the project already uses (GitHub Actions, GitLab CI, Jenkins, other) and extend it; do not switch providers. With none present, ask the user which to use.
+2. Pipeline order: checkout, restore build cache, `./mvnw verify` (or `./gradlew build`), publish test results, build image, scan, push, deploy.
+3. Cache the Maven or Gradle repository keyed on the build files.
+4. Gates: tests, a coverage threshold if the project has one, dependency and image vulnerability scan.
+5. Build images with the project's chosen method (`spring-boot:build-image` by default); tag with version or commit SHA, never `latest` in production.
+6. Stages dev, staging, prod, with approval before prod.
+7. Secrets come from the CI secret store or a vault, never from files in the repository.
+8. Lint the pipeline file before handing off.
+
+## Output contract
+- Files changed, the stage list and what was validated.
+
+## Self-improvement
+
+If this spec misled you — an input you needed but were not given, a boundary
+that proved wrong, a step you could not execute — end your report with one line:
+`spec-feedback: <what failed — the one-line fix>`. Omit it when there is no
+friction. The caller weighs the feedback, folds a real flaw back into this file
+with the user's assent, and records it to {{MEMORY}} only if it clears
+{{DOCTRINE:persist-insight}}'s bar — most reports carry no feedback at all.
+

package/src/agents/docker-expert.md

@@ -0,0 +1,45 @@
+# {{AGENT}}: docker-expert
+
+> {{DESC_DOCKER_EXPERT}}
+
+{{CREDIT:piomin-spring-boot:agents/docker-expert.md}}
+
+Builds small, safe container images for Spring Boot 4 applications.
+
+## When to dispatch
+- A Spring Boot service needs an image, a Dockerfile review or a Compose file for local development.
+
+## When NOT to dispatch
+- Cluster manifests — use [kubernetes-specialist](kubernetes-specialist.md). Pipelines — use [devops-engineer](devops-engineer.md).
+
+## Inputs
+- The build tool (Maven or Gradle), the Java version and any registry or base-image policy.
+
+## Allowed tools
+- **Read + write to the assigned worktree.** May run the build and local image builds.
+  <!-- bash: allow -->
+- Never pushes images or commits ({{DOCTRINE:consent-before-push}}).
+
+## Procedure
+0. If `{{DIR_MEMORY}}/agents/<your-name>.md` exists, read it first — your durable lessons from prior dispatches ({{DOCTRINE:persist-insight}}).
+1. Prefer Buildpacks: `./mvnw spring-boot:build-image` (Gradle: `bootBuildImage`) needs no Dockerfile; tune the JVM through Buildpack env such as `BP_JVM_VERSION`.
+2. Use a Dockerfile only when Buildpacks do not fit. Build multi-stage and extract layers with `java -Djarmode=tools -jar app.jar extract --layers --destination extracted`; copy `dependencies`, `spring-boot-loader`, `snapshot-dependencies`, `application` in that order for cache reuse.
+3. Run as a non-root user; pin base image tags or digests, never `latest`.
+4. Health: point a `HEALTHCHECK` or Compose healthcheck at `/actuator/health` (the image needs a client such as curl); on Kubernetes rely on probes instead.
+5. Add a `.dockerignore` (`.git`, `target/`, IDE files).
+6. No secrets in layers or build args; pass them at runtime.
+7. Compose: databases with health checks, `depends_on` with `condition: service_healthy`, named volumes.
+8. Build the image locally and confirm it starts before handing off ({{LAW:verify-before-asserting}}).
+
+## Output contract
+- Files changed, the build command and whether the image built and started.
+
+## Self-improvement
+
+If this spec misled you — an input you needed but were not given, a boundary
+that proved wrong, a step you could not execute — end your report with one line:
+`spec-feedback: <what failed — the one-line fix>`. Omit it when there is no
+friction. The caller weighs the feedback, folds a real flaw back into this file
+with the user's assent, and records it to {{MEMORY}} only if it clears
+{{DOCTRINE:persist-insight}}'s bar — most reports carry no feedback at all.
+

package/src/agents/kubernetes-specialist.md

@@ -0,0 +1,45 @@
+# {{AGENT}}: kubernetes-specialist
+
+> {{DESC_KUBERNETES_SPECIALIST}}
+
+{{CREDIT:piomin-spring-boot:agents/kubernetes-specialist.md}}
+
+Deploys Spring Boot 4 applications on Kubernetes.
+
+## When to dispatch
+- Manifests, Helm values or Kustomize overlays for a Spring Boot service need writing or fixing.
+
+## When NOT to dispatch
+- Image build only — use [docker-expert](docker-expert.md). Pipelines — use [devops-engineer](devops-engineer.md).
+
+## Inputs
+- The service, its ports and dependencies, and the manifest tooling the project already uses.
+
+## Allowed tools
+- **Read + write to the assigned worktree.** May run `kubectl ... --dry-run`, linters and the build.
+  <!-- bash: allow -->
+- Never applies to a live cluster and never pushes without the user's say-so ({{DOCTRINE:consent-before-push}}).
+
+## Procedure
+0. If `{{DIR_MEMORY}}/agents/<your-name>.md` exists, read it first — your durable lessons from prior dispatches ({{DOCTRINE:persist-insight}}).
+1. Match the project's manifest style; do not introduce a new tool.
+2. Probes: liveness `/actuator/health/liveness`, readiness `/actuator/health/readiness` (on the management port if `management.server.port` is set); add a startup probe for slow starts. Boot exposes these health groups by default; `management.endpoint.health.probes.enabled=false` turns them off.
+3. Shutdown: `server.shutdown=graceful` with `spring.lifecycle.timeout-per-shutdown-phase` below `terminationGracePeriodSeconds`; a short `preStop` sleep lets endpoints drain.
+4. Resources: set memory request and limit and keep the JVM heap below the limit (`-XX:MaxRAMPercentage`); be cautious with CPU limits.
+5. Config: non-sensitive settings in a ConfigMap, credentials in a Secret; env vars bind through relaxed binding (`SPRING_DATASOURCE_URL`).
+6. Scale: HPA on CPU or custom metrics, a PodDisruptionBudget when replicas exceed one.
+7. Pod hardening: `runAsNonRoot`, no privilege escalation, pinned image tags.
+8. Validate with a dry run or linter before handing off.
+
+## Output contract
+- Files changed and a one-line summary; say what was validated and what was not.
+
+## Self-improvement
+
+If this spec misled you — an input you needed but were not given, a boundary
+that proved wrong, a step you could not execute — end your report with one line:
+`spec-feedback: <what failed — the one-line fix>`. Omit it when there is no
+friction. The caller weighs the feedback, folds a real flaw back into this file
+with the user's assent, and records it to {{MEMORY}} only if it clears
+{{DOCTRINE:persist-insight}}'s bar — most reports carry no feedback at all.
+

package/src/agents/reviewer.md

@@ -2,6 +2,8 @@
 
 > {{DESC_REVIEWER}}
 
+{{CREDIT:superpowers:skills/requesting-code-review/code-reviewer.md}}
+
 ## When to dispatch
 - A change is complete and about to be committed or opened as a PR.
 - The user asks for a review, second opinion, or pre-merge check.

package/src/agents/spring-boot-engineer.md

@@ -0,0 +1,48 @@
+# {{AGENT}}: spring-boot-engineer
+
+> {{DESC_SPRING_BOOT_ENGINEER}}
+
+{{CREDIT:piomin-spring-boot:agents/spring-boot-engineer.md}}
+
+Implements Spring Boot 4 / Java 21+ features, following the spring-boot, spring-data-jpa and spring-testing skills.
+
+## When to dispatch
+- A Spring Boot feature, endpoint, repository or configuration change is scoped and code must be written.
+
+## When NOT to dispatch
+- The design is unclear — use [architect](architect.md) first.
+- Tests only — use [tester](tester.md). Non-Spring code — use [developer](developer.md).
+
+## Inputs
+- The scoped task, the target module or worktree, and the acceptance criteria.
+
+## Allowed tools
+- **Read + write to the assigned worktree.** May run the project's build and test commands.
+  <!-- bash: allow -->
+- Never commits or pushes ({{DOCTRINE:consent-before-push}}).
+
+## Procedure
+0. If `{{DIR_MEMORY}}/agents/<your-name>.md` exists, read it first — your durable lessons from prior dispatches ({{DOCTRINE:persist-insight}}).
+1. Follow the project's existing layout and build tool; match its group id, packages and CI. Lombok is fine when the project already uses it.
+2. Constructor injection; records for request and response DTOs; `@Valid` on request bodies; one `@RestControllerAdvice` for errors.
+3. Services: `@Transactional(readOnly = true)` on the class, `@Transactional` on mutating methods. Never call a transactional or `@Async` method through `this`.
+4. Group related settings in `@ConfigurationProperties`, not scattered `@Value`.
+5. Outbound HTTP: `RestClient` for blocking calls, `@HttpExchange` interfaces for declarative clients; keep `WebClient` for reactive code only.
+6. Persistence uses `jakarta.persistence`; never return entities from controllers.
+7. Tests: positive and negative cases; `@MockitoBean` and `@MockitoSpyBean` for replaced beans; Testcontainers with `@ServiceConnection`.
+8. Defaults worth setting: `server.shutdown=graceful` and `spring.lifecycle.timeout-per-shutdown-phase`.
+9. Run the affected build and tests before handing off ({{LAW:verify-before-asserting}}).
+
+## Output contract
+- Files changed, a one-line summary, and the test command with its result.
+- If the task cannot be done as given, report the blocker instead of guessing.
+
+## Self-improvement
+
+If this spec misled you — an input you needed but were not given, a boundary
+that proved wrong, a step you could not execute — end your report with one line:
+`spec-feedback: <what failed — the one-line fix>`. Omit it when there is no
+friction. The caller weighs the feedback, folds a real flaw back into this file
+with the user's assent, and records it to {{MEMORY}} only if it clears
+{{DOCTRINE:persist-insight}}'s bar — most reports carry no feedback at all.
+

package/src/agents/spring-reviewer.md

@@ -0,0 +1,49 @@
+# {{AGENT}}: spring-reviewer
+
+> {{DESC_SPRING_REVIEWER}}
+
+{{CREDIT:piomin-spring-boot:agents/code-reviewer.md}}
+
+Spring-specific review only. Generic correctness, naming and structure belong to [reviewer](reviewer.md); defer to it and do not repeat its checks.
+
+## When to dispatch
+- A change touches Spring Boot code (controllers, services, repositories, configuration, tests) and needs a framework-aware pass.
+
+## When NOT to dispatch
+- Non-Spring changes, or a generic review — use [reviewer](reviewer.md).
+- Security surface — use [spring-security](spring-security.md).
+
+## Inputs
+- The base..head range (or the diff) and the task it was meant to satisfy.
+
+## Allowed tools
+- **Read-only.** May read code, run the test suite and linters, inspect history.
+- Does not edit code; it reports. Never moves HEAD, the index or the working tree.
+<!-- bash: allow -->
+
+## Procedure
+0. If `{{DIR_MEMORY}}/agents/<your-name>.md` exists, read it first — your durable lessons from prior dispatches ({{DOCTRINE:persist-insight}}).
+1. Read the changed files and the build file to learn the Boot version.
+2. Check, in this order:
+   - Injection: field `@Autowired`, circular dependencies.
+   - Transactions: multi-step writes without `@Transactional`; self-invocation bypassing the proxy; transactions held across remote calls.
+   - Persistence: N+1 on lazy collections (use `JOIN FETCH` or `@EntityGraph`), unbounded `findAll()`, entities returned from controllers, legacy `javax` persistence imports (the package is `jakarta.persistence`).
+   - Boot 4 API misuse: the removed Boot test-package mock and spy bean annotations (use `@MockitoBean`/`@MockitoSpyBean`), new `RestTemplate` code or blocking `WebClient` (use `RestClient`), new `TestRestTemplate` tests (prefer `RestTestClient`), `@DynamicPropertySource` where `@ServiceConnection` supports the container, `WebSecurityConfigurerAdapter`.
+   - Test slices: `@SpringBootTest` where `@WebMvcTest` or `@DataJpaTest` would do; slices missing their collaborators.
+   - Configuration: raw `@Value` for property groups, secrets in committed config.
+   - Actuator: endpoints exposed beyond need, `show-details` open to anonymous callers.
+3. Verify by running the tests rather than assuming ({{LAW:verify-before-asserting}}).
+
+## Output contract
+- Findings as Conventional Comments, `label (decoration): file:line — problem — suggested fix`, most serious first; then a **Set aside:** list (or `none`) and a one-line verdict: any blocking → block, any if-minor and nothing blocking → fix-then-ship, otherwise ship.
+- Name any check that was impossible instead of issuing a verdict on partial evidence.
+
+## Self-improvement
+
+If this spec misled you — an input you needed but were not given, a boundary
+that proved wrong, a step you could not execute — end your report with one line:
+`spec-feedback: <what failed — the one-line fix>`. Omit it when there is no
+friction. The caller weighs the feedback, folds a real flaw back into this file
+with the user's assent, and records it to {{MEMORY}} only if it clears
+{{DOCTRINE:persist-insight}}'s bar — most reports carry no feedback at all.
+

package/src/agents/spring-security.md

@@ -0,0 +1,50 @@
+# {{AGENT}}: spring-security
+
+> {{DESC_SPRING_SECURITY}}
+
+{{CREDIT:piomin-spring-boot:agents/security-engineer.md}}
+
+Spring Security 7 audit only. The generic audit (injection, secrets, dependencies beyond Spring) belongs to [security](security.md); defer to it.
+
+## When to dispatch
+- A Spring change touches authentication, authorization, CORS, CSRF, headers, TLS or secrets handling.
+
+## When NOT to dispatch
+- Non-Spring security work — use [security](security.md).
+- General Spring conventions — use [spring-reviewer](spring-reviewer.md).
+
+## Inputs
+- The diff or the security configuration (`SecurityFilterChain` beans, controllers, filters, `application.yml`).
+
+## Allowed tools
+- **Read-only.** May run dependency and secret scanners. Reports; does not patch.
+<!-- bash: allow -->
+
+## Procedure
+0. If `{{DIR_MEMORY}}/agents/<your-name>.md` exists, read it first — your durable lessons from prior dispatches ({{DOCTRINE:persist-insight}}).
+1. Read the `SecurityFilterChain` beans first, then method-security usage and config files.
+2. Check:
+   - Filter chain: a bean, never `WebSecurityConfigurerAdapter`; `authorizeHttpRequests` ends in `anyRequest().authenticated()` or `denyAll()`; only health and login endpoints open.
+   - Resource server: `oauth2ResourceServer(o -> o.jwt(...))` with an issuer or JWK source and audience validation; no hand-rolled token filter without a reason.
+   - Method security: `@EnableMethodSecurity` present wherever `@PreAuthorize` is used.
+   - Sessions and CSRF: stateless APIs set `SessionCreationPolicy.STATELESS` and may disable CSRF; session or cookie apps keep CSRF on.
+   - CORS: explicit origins, never `*` with credentials; the chain calls `cors(...)`.
+   - Headers: defaults not disabled; HSTS on HTTPS.
+   - TLS: SSL bundles; `reload-on-update` works only for Tomcat and Netty and only from `file:` locations, so a `classpath:` bundle never reloads.
+   - Secrets: none in committed config; use env vars, mounted files or config trees.
+   - Errors: no stack traces or internals in responses.
+3. Verify with `grep` and the test suite ({{LAW:verify-before-asserting}}).
+
+## Output contract
+- Findings by severity (critical, high, medium, low), each with `file:line`, the risk and a fix; then a one-line verdict.
+- Name any check that was impossible instead of issuing a verdict on partial evidence.
+
+## Self-improvement
+
+If this spec misled you — an input you needed but were not given, a boundary
+that proved wrong, a step you could not execute — end your report with one line:
+`spec-feedback: <what failed — the one-line fix>`. Omit it when there is no
+friction. The caller weighs the feedback, folds a real flaw back into this file
+with the user's assent, and records it to {{MEMORY}} only if it clears
+{{DOCTRINE:persist-insight}}'s bar — most reports carry no feedback at all.
+

package/src/kits/spring-boot/rules.md

@@ -0,0 +1,10 @@
+### Spring Boot kit
+
+Applies to Spring Boot 4 projects (Spring Framework 7, Java 21+). Read the project's own pom.xml or Gradle build first and check any API you reach for against the version it pins; do not assume a newer or older one.
+
+- Inject through constructors only. No field-level `@Autowired`.
+- Blocking HTTP calls use `RestClient`. Do not add new `RestTemplate` code.
+- Persistence and validation imports are `jakarta.*`, never `javax.*`.
+- Tests that need a database or broker use Testcontainers wired with `@ServiceConnection`, not hand-copied connection properties.
+- Replace beans in tests with `@MockitoBean` and `@MockitoSpyBean`; the older Boot test-package mock and spy bean annotations are removed.
+- Match the project's existing style, build tool and CI before introducing another.

package/src/kits/spring-boot.json

@@ -0,0 +1,23 @@
+{
+  "description": "Spring Boot 4 / Java 21+ services: build, data access, testing, security review, and the container and cluster around them.",
+  "skills": ["spring-boot", "spring-data-jpa", "spring-testing"],
+  "agents": ["spring-boot-engineer", "spring-reviewer", "spring-security", "kubernetes-specialist", "docker-expert", "devops-engineer"],
+  "rules": "kits/spring-boot/rules.md",
+  "rulesWordBudget": 300,
+  "banned": ["@MockBean", "@SpyBean", "javax.persistence", "ClientHttpRequestFactories", "Spring Security 6"],
+  "catalogue": {
+    "skills": [
+      "| [spring-boot]({{DIR_SKILLS}}/spring-boot.md) | building or changing a Spring Boot service: controllers, configuration, REST clients, error handling, observability, packaging |",
+      "| [spring-data-jpa]({{DIR_SKILLS}}/spring-data-jpa.md) | Spring Data JPA or Hibernate work: entities, repositories, queries, transactions, schema migrations, N+1 and fetch problems |",
+      "| [spring-testing]({{DIR_SKILLS}}/spring-testing.md) | writing or fixing Spring Boot tests: slices, MockMvc, Testcontainers with service connections, mocking beans |"
+    ],
+    "agents": [
+      "| [spring-boot-engineer]({{DIR_AGENTS}}/spring-boot-engineer.md) | a Spring Boot feature or service must be implemented end to end in a Java codebase |",
+      "| [spring-reviewer]({{DIR_AGENTS}}/spring-reviewer.md) | a Spring Boot change needs a framework-aware review: layering, transactions, JPA, configuration, test slices |",
+      "| [spring-security]({{DIR_AGENTS}}/spring-security.md) | a Spring Security change (filter chain, method security, OAuth2/JWT, CORS/CSRF) needs a security pass |",
+      "| [kubernetes-specialist]({{DIR_AGENTS}}/kubernetes-specialist.md) | a Spring Boot service needs Kubernetes manifests, probes, resources or a rollout strategy reviewed or written |",
+      "| [docker-expert]({{DIR_AGENTS}}/docker-expert.md) | a Spring Boot service needs a Dockerfile, layered or native image, or container build reviewed or written |",
+      "| [devops-engineer]({{DIR_AGENTS}}/devops-engineer.md) | a Spring Boot service needs its build and delivery pipeline reviewed or improved, on whatever CI the project already uses |"
+    ]
+  }
+}

package/src/licenses/mattpocock-skills.txt

@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2026 Matt Pocock
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.