diff --git a/README.md b/README.md
index dfc62b1..fab23bb 100644
--- a/README.md
+++ b/README.md
@@ -7,7 +7,8 @@ surfaces.
 ## Current status
 
 Phase 3 is complete, and Phase 4A (Core identity foundation), Phase 4B (identity HTTP transport),
-and Phase 4C (Core authorization) have landed. The repository includes
+Phase 4C (Core authorization), and Phase 5A (module authorization integration) have landed. The
+repository includes
 public module contracts, independently versioned capabilities, static module manifests,
 deterministic automatic discovery, compiled server loading, a staged capability registry,
 dependency-ordered lifecycle with rollback, module-owned Drizzle table contracts, deterministic
@@ -46,11 +47,26 @@ immutable permission registry; an explicit, single-use, concurrency-safe platfor
 command (`pnpm authorization:bootstrap-owner`, never HTTP-exposed, never inferred from signup
 order); and, in `apps/api`, reusable Hono authorization enforcement plus a narrowly scoped
 `/api/v1/grants` HTTP API for managing *explicit* grants only, requiring an exact trusted `Origin`
-on every mutation. Roles/groups/teams/tenants, an owner-management HTTP endpoint, service
-accounts/API tokens, and module-permission registration remain later work. See
-[Authorization](docs/development/authorization.md) and
+on every mutation. Roles/groups/teams/tenants, an owner-management HTTP endpoint, and service
+accounts/API tokens remain later work. See [Authorization](docs/development/authorization.md) and
 [ADR 0013](docs/architecture/decisions/0013-core-authorization-and-owner-bootstrap.md).
 
+Phase 5A wires module-declared permissions into that same authorization system: a version-3 module
+manifest declares permissions as complete `{id, baseline, description}` declarations (never a bare
+string), and `apps/api` discovers every installed module and composes one immutable platform
+permission registry — Core's own declarations plus every module's — before constructing its
+database pool, Better Auth instance, or HTTP server. A module's declared permission is then
+enforceable and grantable through the exact same `requirePermission` middleware and
+`/api/v1/grants` HTTP API Core's own permissions already use, with `core.*` reserved and never
+claimable by a module. `pnpm doctor` and `pnpm modules:validate` both validate permission
+composition. Installing, removing, or editing a module currently requires an API restart; removing
+one makes its permission unavailable to new checks and new grants without deleting or revoking any
+historical grant or audit row. Module HTTP route registration and relationship-derived grant
+creation remain later work. See
+[Authorization](docs/development/authorization.md#module-permission-composition),
+[Module development](docs/module-development/README.md), and
+[ADR 0014](docs/architecture/decisions/0014-module-permission-declarations-and-platform-registry.md).
+
 ## Quick start
 
 Prerequisites: Node.js 24 LTS, pnpm 11.12, and Docker with Compose.
@@ -89,6 +105,8 @@ Modules
     0 modules installed
   ✓ Dependency graph valid
     Dependency order is empty.
+  ✓ Permission registry composed
+    1 permission registered
 
 Data definitions
   ✓ Schemas validated
diff --git a/apps/api/src/app.ts b/apps/api/src/app.ts
index 8acd5fa..fce51be 100644
--- a/apps/api/src/app.ts
+++ b/apps/api/src/app.ts
@@ -1,8 +1,8 @@
 import { isAuthenticatedPrincipal } from "@ekklesia/contracts";
-import {
-  type AuthorizationEvaluator,
-  type AuthorizationStore,
-  CORE_PERMISSION_REGISTRY,
+import type {
+  AuthorizationEvaluator,
+  AuthorizationStore,
+  PermissionRegistry,
 } from "@ekklesia/core/authorization";
 import { type EkklesiaIdentityRuntime, IdentityRuntimeError } from "@ekklesia/core/identity";
 import { Hono } from "hono";
@@ -12,6 +12,10 @@ import { createGrantsRoutes } from "./grants-routes.js";
 
 export interface CreateApiAppAuthorizationOptions {
   readonly evaluator: AuthorizationEvaluator;
+  /** The complete platform permission registry (Core's own declarations plus every installed
+   * module's) — always the caller-composed registry `startApi` produces via
+   * `composeStartupPermissionRegistry`, never a Core-only fallback constructed here. */
+  readonly registry: PermissionRegistry;
   readonly store: AuthorizationStore;
 }
 
@@ -80,7 +84,7 @@ export function createApiApp(options: CreateApiAppOptions): Hono {
     "/",
     createGrantsRoutes({
       evaluator: options.authorization.evaluator,
-      registry: CORE_PERMISSION_REGISTRY,
+      registry: options.authorization.registry,
       runtime,
       store: options.authorization.store,
       trustedOrigins: allowedOrigins,
diff --git a/apps/api/src/authorization-middleware.ts b/apps/api/src/authorization-middleware.ts
index 813d184..9b88124 100644
--- a/apps/api/src/authorization-middleware.ts
+++ b/apps/api/src/authorization-middleware.ts
@@ -17,9 +17,11 @@ function sanitizedDenial(c: Context, status: 403 | 503, code: string): Response
 }
 
 /**
- * Builds a Hono middleware that enforces `permissionId`, resolved exclusively from Core's own
- * trusted, server-side permission registry — a route never supplies its own baseline, and no
- * request input can influence which baseline is checked. The principal is resolved exclusively
+ * Builds a Hono middleware that enforces `permissionId`, resolved exclusively from the injected,
+ * trusted, server-side `PermissionRegistry` — in production, the complete platform registry
+ * `startApi` composes from Core's own declarations plus every installed module's, never a
+ * Core-only fallback. A route never supplies its own baseline, and no request input can influence
+ * which baseline is checked. The principal is resolved exclusively
  * through the injected identity runtime (`resolvePrincipal`), the same one `/api/v1/principal`
  * already uses; nothing here ever trusts a client-supplied claim about identity, ownership, or
  * grants. A permission id that is not registered fails closed (403) exactly like a denied
diff --git a/apps/api/src/platform-permissions.ts b/apps/api/src/platform-permissions.ts
new file mode 100644
index 0000000..707a38e
--- /dev/null
+++ b/apps/api/src/platform-permissions.ts
@@ -0,0 +1,106 @@
+import { readFile } from "node:fs/promises";
+import path from "node:path";
+import { fileURLToPath } from "node:url";
+import type { PermissionRegistry } from "@ekklesia/core/authorization";
+import {
+  composePlatformPermissionRegistry,
+  discoverModules,
+  type DiscoveredModule,
+} from "@ekklesia/core/modules";
+
+export type PlatformPermissionConfigurationErrorCode = "PLATFORM_PERMISSION_CONFIGURATION_INVALID";
+
+/** Never carries the underlying package.json path, filesystem error, module directory, manifest
+ * content, or any other machine detail — a broken repository layout is a startup-time
+ * configuration problem, not something to describe in detail to whatever eventually logs it. */
+export class PlatformPermissionConfigurationError extends Error {
+  readonly code: PlatformPermissionConfigurationErrorCode;
+
+  constructor() {
+    super("Module discovery configuration could not be resolved.");
+    this.name = "PlatformPermissionConfigurationError";
+    this.code = "PLATFORM_PERMISSION_CONFIGURATION_INVALID";
+    this.stack = `${this.name}: ${this.message}`;
+    Object.freeze(this);
+  }
+}
+
+/**
+ * Resolves the repository root from this compiled/source file's own location, never
+ * `process.cwd()` — package-scoped test runs execute with `apps/api` as the working directory,
+ * which would silently point discovery at the wrong `modules/` directory. `apps/api/dist/*.js`
+ * (compiled) and `apps/api/src/*.ts` (source, under a test runner transforming TypeScript
+ * on the fly) sit at the identical depth from the repository root, so the same fixed number of
+ * `..` segments resolves correctly in both.
+ */
+function resolveRepositoryRoot(): string {
+  return path.resolve(path.dirname(fileURLToPath(import.meta.url)), "..", "..", "..");
+}
+
+interface PackageVersionFile {
+  readonly version?: unknown;
+}
+
+async function readPackageVersion(packagePath: string): Promise<string> {
+  let packageData: PackageVersionFile;
+  try {
+    packageData = JSON.parse(await readFile(packagePath, "utf8")) as PackageVersionFile;
+  } catch {
+    throw new PlatformPermissionConfigurationError();
+  }
+  if (typeof packageData.version !== "string" || packageData.version.trim().length === 0) {
+    throw new PlatformPermissionConfigurationError();
+  }
+  return packageData.version;
+}
+
+export interface ComposeStartupPermissionRegistryOptions {
+  /** Test-only escape hatch: overrides the directory module discovery scans for installed
+   * modules. Never set by the executable entry point — production always discovers the real
+   * repository's `modules/` directory. Platform and Module SDK versions are always read from the
+   * real repository's own `package.json` files regardless of this override. */
+  readonly modulesDirectoryOverride?: string;
+  /** Test-only escape hatch: supplies an already-discovered module list directly, skipping the
+   * filesystem discovery step entirely, while still running the real `composePlatformPermissionRegistry`
+   * on it. Exists because a genuine exact permission-id collision between two modules is
+   * unreachable through real discovery: module ids are unique (discovery itself rejects a
+   * duplicate) and every permission id is namespaced by its own module's id, so no two
+   * really-discovered modules can ever contribute the same permission id. Proving
+   * `PermissionRegistryError`'s own duplicate-rejection behavior in composition, and that startup
+   * has created no pool before it runs, requires supplying deliberately conflicting declarations
+   * directly. (A module-vs-Core collision, by contrast, *is* reachable through real discovery — a
+   * module whose own id is `"core"` — and surfaces as `ModulePermissionCompositionError`, not this
+   * case; `modulesDirectoryOverride` with a real fixture is what proves that one.) Never set by the
+   * executable entry point.
+   */
+  readonly discoveredModulesOverride?: readonly DiscoveredModule[];
+}
+
+/**
+ * Discovers every installed module's manifest and composes the one immutable platform permission
+ * registry (Core's own declarations plus every module's) — the production source of truth
+ * `apps/api` enforces every request against. Only manifests are read; no module server entrypoint
+ * is imported or executed, and no module lifecycle hook runs. Read the platform and Module SDK
+ * versions only when this function actually executes, never at import time.
+ */
+export async function composeStartupPermissionRegistry(
+  options: ComposeStartupPermissionRegistryOptions = {},
+): Promise<PermissionRegistry> {
+  if (options.discoveredModulesOverride) {
+    return composePlatformPermissionRegistry(options.discoveredModulesOverride);
+  }
+
+  const repositoryRoot = resolveRepositoryRoot();
+  const [platformVersion, moduleSdkVersion] = await Promise.all([
+    readPackageVersion(path.join(repositoryRoot, "package.json")),
+    readPackageVersion(path.join(repositoryRoot, "packages/module-sdk/package.json")),
+  ]);
+
+  const modules = await discoverModules({
+    moduleSdkVersion,
+    modulesDirectory: options.modulesDirectoryOverride ?? path.join(repositoryRoot, "modules"),
+    platformVersion,
+  });
+
+  return composePlatformPermissionRegistry(modules);
+}
diff --git a/apps/api/src/server.ts b/apps/api/src/server.ts
index c4b2760..754df1b 100644
--- a/apps/api/src/server.ts
+++ b/apps/api/src/server.ts
@@ -5,11 +5,13 @@ import {
   createAuthorizationStore,
 } from "@ekklesia/core/authorization";
 import { createIdentityRuntime, type IdentityRateLimitOverride } from "@ekklesia/core/identity";
+import type { DiscoveredModule } from "@ekklesia/core/modules";
 import { serve, type ServerType } from "@hono/node-server";
 import type { Hono } from "hono";
 import { createApiApp } from "./app.js";
 import { type ApiRuntimeConfig, createApiRuntimeConfig } from "./config.js";
 import { type ApiDatabaseHandle, createApiDatabase } from "./database.js";
+import { composeStartupPermissionRegistry } from "./platform-permissions.js";
 
 const connectionTimeoutMs = 10_000;
 const operationTimeoutMs = 10_000;
@@ -24,6 +26,13 @@ export interface StartApiOptions {
   /** Test-only escape hatch: never set by the executable entry point. See
    * `IdentityRateLimitOverride` for why production must leave this unset. */
   readonly rateLimitOverride?: IdentityRateLimitOverride;
+  /** Test-only escape hatch: overrides the directory permission-composition module discovery
+   * scans. Never set by the executable entry point. See `composeStartupPermissionRegistry`. */
+  readonly modulesDirectoryOverride?: string;
+  /** Test-only escape hatch: supplies an already-discovered module list directly to permission
+   * composition, skipping filesystem discovery entirely. Never set by the executable entry
+   * point. See `composeStartupPermissionRegistry`. */
+  readonly discoveredModulesOverride?: readonly DiscoveredModule[];
 }
 
 async function safeCloseDatabase(handle: ApiDatabaseHandle): Promise<void> {
@@ -54,18 +63,33 @@ function bindServer(app: Hono, host: string, port: number): Promise<ServerType>
 }
 
 /**
- * Explicit runtime composition: constructs the one shared PostgreSQL pool (identity and
- * authorization each get a typed Drizzle view over it — see `./database.js`), the Core identity
- * runtime, the Core authorization store and evaluator, and the Hono application, then starts the
- * Node HTTP server. Returns a handle with an idempotent `shutdown()` so both the executable entry
- * point and tests can stop the server and close the shared pool deterministically. Never applies
- * migrations or generates schema — the database must already be migrated through
- * `pnpm migrations:apply`.
+ * Explicit runtime composition, in a fixed order: discover every installed module's manifest and
+ * compose the one immutable platform permission registry from Core's own declarations plus every
+ * module's (`./platform-permissions.js`); only then construct the one shared PostgreSQL pool
+ * (identity and authorization each get a typed Drizzle view over it — see `./database.js`), the
+ * Core identity runtime, the Core authorization store and evaluator, and the Hono application;
+ * finally start the Node HTTP server. Permission composition happens before any pool, Better
+ * Auth instance, or HTTP server exists, so a discovery or composition failure — an invalid
+ * manifest, a dependency-graph problem, a module claiming the reserved `core.*` permission
+ * namespace, or a duplicate permission declaration — leaves nothing to close: it propagates
+ * directly, and startup never falls back to a Core-only registry. Returns a handle with an
+ * idempotent `shutdown()` so both the executable entry point and tests can stop the server and
+ * close the shared pool deterministically. Never applies migrations or generates schema — the
+ * database must already be migrated through `pnpm migrations:apply`.
  */
 export async function startApi(
   config: ApiRuntimeConfig,
   options: StartApiOptions = {},
 ): Promise<RunningApi> {
+  const permissionRegistry = await composeStartupPermissionRegistry({
+    ...(options.modulesDirectoryOverride === undefined
+      ? {}
+      : { modulesDirectoryOverride: options.modulesDirectoryOverride }),
+    ...(options.discoveredModulesOverride === undefined
+      ? {}
+      : { discoveredModulesOverride: options.discoveredModulesOverride }),
+  });
+
   const databaseHandle = createApiDatabase(config.databaseUrl, {
     connectionTimeoutMs,
     operationTimeoutMs,
@@ -91,7 +115,7 @@ export async function startApi(
   });
 
   const app = createApiApp({
-    authorization: { evaluator, store: authorizationStore },
+    authorization: { evaluator, registry: permissionRegistry, store: authorizationStore },
     runtime,
     trustedOrigins: config.identity.trustedOrigins,
   });
diff --git a/apps/api/tests/api-http.integration.ts b/apps/api/tests/api-http.integration.ts
index c997385..902ae1b 100644
--- a/apps/api/tests/api-http.integration.ts
+++ b/apps/api/tests/api-http.integration.ts
@@ -8,12 +8,13 @@ import {
   createAuthorizationDrizzleDatabase,
   createAuthorizationStore,
 } from "@ekklesia/core/authorization";
+import { diagnoseEkklesia } from "@ekklesia/core/diagnostics";
 import {
   applyModuleMigrations,
   createPostgresMigrationDatabase,
   discoverAllMigrations,
 } from "@ekklesia/core/migrations";
-import { mkdtemp, rm } from "node:fs/promises";
+import { mkdir, mkdtemp, rm, writeFile } from "node:fs/promises";
 import { tmpdir } from "node:os";
 import { Client, type QueryResultRow } from "pg";
 import { afterAll, describe, expect, it, vi } from "vitest";
@@ -198,10 +199,60 @@ async function applyCoreMigrations(url: string): Promise<void> {
   }
 }
 
+const trackedModulesDirectories: string[] = [];
+
+async function createFixtureModulesDirectory(): Promise<string> {
+  const directory = await mkdtemp(path.join(tmpdir(), "ekklesia-api-installed-modules-"));
+  trackedModulesDirectories.push(directory);
+  return directory;
+}
+
+/** Writes a minimal, structurally valid module fixture straight to disk — real discovery and
+ * composition run against it, nothing hand-built. The server entrypoint throws immediately if
+ * ever imported/executed, so a scenario that completes without that throw firing is direct
+ * evidence permission composition and Doctor never ran module lifecycle code. */
+async function createFixtureModule(
+  modulesDirectory: string,
+  id: string,
+  manifestOverrides: Record<string, unknown> = {},
+): Promise<void> {
+  const moduleDirectory = path.join(modulesDirectory, id);
+  const manifest = {
+    manifestVersion: 3,
+    id,
+    name: `${id} module`,
+    version: "0.1.0",
+    compatibility: { platform: "^0.1.0", moduleSdk: "^0.1.0" },
+    permissions: [],
+    entrypoints: { server: "./dist/server/index.js" },
+    ...manifestOverrides,
+  };
+  await mkdir(path.join(moduleDirectory, "dist/server"), { recursive: true });
+  await Promise.all([
+    writeFile(
+      path.join(moduleDirectory, "package.json"),
+      JSON.stringify({
+        ekklesia: { layer: "module" },
+        name: `@test/module-${id}`,
+        type: "module",
+        version: "0.1.0",
+      }),
+    ),
+    writeFile(path.join(moduleDirectory, "ekklesia.module.json"), JSON.stringify(manifest)),
+    writeFile(
+      path.join(moduleDirectory, "dist/server/index.js"),
+      'throw new Error("permission composition and Doctor must never execute a module server entrypoint");',
+    ),
+  ]);
+}
+
 afterAll(async () => {
   for (const databaseName of [...trackedDatabases]) {
     await dropDatabase(databaseName).catch(() => undefined);
   }
+  for (const directory of trackedModulesDirectories.splice(0)) {
+    await rm(directory, { recursive: true, force: true }).catch(() => undefined);
+  }
 });
 
 const identityBaseUrl = "https://identity.example.test";
@@ -216,7 +267,10 @@ function apiOrigin(running: RunningApi): string {
   return `http://127.0.0.1:${address.port}`;
 }
 
-async function startMigratedApi(url: string): Promise<RunningApi> {
+async function startMigratedApi(
+  url: string,
+  modulesDirectoryOverride?: string,
+): Promise<RunningApi> {
   const config = createApiRuntimeConfig(
     {
       databaseUrl: url,
@@ -230,10 +284,15 @@ async function startMigratedApi(url: string): Promise<RunningApi> {
     },
     { environment: "production" },
   );
-  // A narrow, explicit test-only override: production composition (server.ts's `main`) never
-  // sets this. Better Auth's built-in sign-in/sign-up rate limit (3 requests / 10s) would
-  // otherwise reject this suite's many sequential requests.
-  return startApi(config, { rateLimitOverride: { max: 1000, window: 1 } });
+  // Narrow, explicit test-only overrides: production composition (server.ts's `main`) never sets
+  // either. Better Auth's built-in sign-in/sign-up rate limit (3 requests / 10s) would otherwise
+  // reject this suite's many sequential requests; `modulesDirectoryOverride` lets a scenario point
+  // permission composition at a real, disposable module fixture instead of this repository's own
+  // (empty) `modules/` directory.
+  return startApi(config, {
+    rateLimitOverride: { max: 1000, window: 1 },
+    ...(modulesDirectoryOverride === undefined ? {} : { modulesDirectoryOverride }),
+  });
 }
 
 function setCookieValue(response: Response): string | undefined {
@@ -832,9 +891,11 @@ describe.sequential("live API — platform owner and delegated grant management"
   const ownerEmail = "authz-owner@example.test";
   const managerEmail = "authz-manager@example.test";
   const password = "correct horse battery staple";
-  // Module permission registration does not exist yet, so `core.authorization.grants.manage` —
-  // the one permission `CORE_PERMISSION_REGISTRY` actually declares — stands in for "some
-  // registered permission" throughout this scenario's explicit-grant lifecycle.
+  // This scenario intentionally proves out the explicit-grant lifecycle (create, list, use,
+  // revoke) against Core's own grant-management permission rather than a module's — a module
+  // permission would exercise the exact same route and evaluator code, adding coverage breadth
+  // without coverage depth. The "live API — installed module permission end-to-end" scenario
+  // below is where a real installed module's declared permission is proven end-to-end instead.
   const testPermission = "core.authorization.grants.manage";
   // Well-formed (passes `PermissionId` shape validation) but never registered — used to prove the
   // live route rejects it even for a caller who otherwise has full grant-management authority.
@@ -1147,3 +1208,198 @@ describe.sequential("live API — authorization infrastructure failure is never
     await running.shutdown().catch(() => undefined);
   });
 });
+
+describe.sequential("live API — installed module permission end-to-end", () => {
+  let database: { name: string; url: string };
+  let modulesDirectory: string;
+  let running: RunningApi;
+  let origin: string;
+  let ownerUserId: string;
+  let ownerCookie: string;
+  let createdGrantId: string;
+
+  const ownerEmail = "authz-module-owner@example.test";
+  const password = "correct horse battery staple";
+  const modulePermission = "calendar.events.read";
+  const absentPermission = "scheduler.jobs.read";
+
+  it("recognizes a real installed module and passes permission composition in Doctor, without executing its server entrypoint", async () => {
+    modulesDirectory = await createFixtureModulesDirectory();
+    await createFixtureModule(modulesDirectory, "calendar", {
+      permissions: [
+        {
+          baseline: "restricted",
+          description: "Read calendar events.",
+          id: modulePermission,
+        },
+      ],
+    });
+
+    const report = await diagnoseEkklesia({
+      moduleDiscovery: {
+        moduleSdkVersion: "0.1.0",
+        modulesDirectory,
+        platformVersion: "0.1.0",
+      },
+      runtimeRange: ">=0.0.0",
+      runtimeVersion: process.versions.node,
+    });
+
+    // Only the module-discovery and permission-composition checks are asserted here — this
+    // fixture's temporary directory has no `core/drizzle` sibling for migrations discovery to
+    // find, which is unrelated to what this scenario proves and would otherwise make the check
+    // fail for a reason that has nothing to do with permission composition.
+    const moduleCheck = report.checks.find((entry) => entry.id === "modules.discovery");
+    const permissionsCheck = report.checks.find((entry) => entry.id === "modules.permissions");
+    expect(moduleCheck).toMatchObject({ status: "pass" });
+    expect(moduleCheck?.details).toEqual(["calendar 0.1.0"]);
+    expect(permissionsCheck).toMatchObject({
+      details: ["2 permissions registered"],
+      status: "pass",
+    });
+  });
+
+  it("starts a freshly migrated API that discovers the fixture module and signs up its eventual owner", async () => {
+    database = await createDisposableDatabase("authz_module");
+    await applyCoreMigrations(database.url);
+    running = await startMigratedApi(database.url, modulesDirectory);
+    origin = apiOrigin(running);
+
+    const ownerSignUp = await signUp(origin, ownerEmail, password);
+    ownerUserId = ownerSignUp.body.user.id;
+    const ownerSignIn = await signIn(origin, ownerEmail, password);
+    expect(ownerSignIn.cookiePair).toBeDefined();
+    ownerCookie = ownerSignIn.cookiePair as string;
+
+    // Existing identity HTTP behavior — sign-up, sign-in, principal resolution — is unaffected by
+    // module permission composition running before it.
+    const { body: principal } = await fetchPrincipal(origin, { cookie: ownerCookie });
+    expect(principal).toEqual({ kind: "authenticated", userId: ownerUserId });
+  });
+
+  it("lets an authorized platform owner create an explicit grant for the installed module's permission, with an atomic audit entry using its declared id", async () => {
+    const { close, db } = createAuthorizationDrizzleDatabase(database.url);
+    try {
+      await bootstrapPlatformOwner(db, { userId: parseEkklesiaUserId(ownerUserId) });
+    } finally {
+      await close();
+    }
+
+    const response = await fetch(`${origin}/api/v1/grants`, {
+      body: JSON.stringify({ permission: modulePermission, subjectUserId: ownerUserId }),
+      headers: {
+        "content-type": "application/json",
+        cookie: ownerCookie,
+        origin: dashboardOrigin,
+      },
+      method: "POST",
+    });
+    const body = (await response.json()) as {
+      grant: { id: string; permission: string; sourceKind: string };
+    };
+
+    expect(response.status).toBe(201);
+    expect(body.grant.permission).toBe(modulePermission);
+    expect(body.grant.sourceKind).toBe("explicit");
+    createdGrantId = body.grant.id;
+
+    const admin = await openAdministrativeClient(database.url);
+    try {
+      const rows = await safeQuery<{ action: string; permission_id: string }>(
+        admin,
+        `select action, permission_id from "ekklesia_core_authorization_audit" where grant_id = $1`,
+        [createdGrantId],
+      );
+      expect(rows).toEqual([{ action: "grant.create", permission_id: modulePermission }]);
+    } finally {
+      await safeEnd(admin);
+    }
+  });
+
+  it("rejects a well-formed permission absent from the composed registry, writing no grant or audit row", async () => {
+    const admin = await openAdministrativeClient(database.url);
+    try {
+      const before = await safeQuery<{ count: string }>(
+        admin,
+        `select count(*)::text as count from "ekklesia_core_grant" where permission_id = $1`,
+        [absentPermission],
+      );
+      expect(before[0]?.count).toBe("0");
+
+      const response = await fetch(`${origin}/api/v1/grants`, {
+        body: JSON.stringify({ permission: absentPermission, subjectUserId: ownerUserId }),
+        headers: {
+          "content-type": "application/json",
+          cookie: ownerCookie,
+          origin: dashboardOrigin,
+        },
+        method: "POST",
+      });
+      expect(response.status).toBe(403);
+      expect(await response.json()).toEqual({ code: "AUTHORIZATION_DENIED" });
+
+      const after = await safeQuery<{ count: string }>(
+        admin,
+        `select count(*)::text as count from "ekklesia_core_grant" where permission_id = $1`,
+        [absentPermission],
+      );
+      const audit = await safeQuery<{ count: string }>(
+        admin,
+        `select count(*)::text as count from "ekklesia_core_authorization_audit" where permission_id = $1`,
+        [absentPermission],
+      );
+      expect(after[0]?.count).toBe("0");
+      expect(audit[0]?.count).toBe("0");
+    } finally {
+      await safeEnd(admin);
+    }
+  });
+
+  it("shuts down cleanly, then restarting with the module absent rejects new grants for its former permission while the historical grant remains listable", async () => {
+    await running.shutdown();
+    const remainingConnections = await activeConnectionCount(database.name);
+    expect(remainingConnections).toBe(0);
+
+    const emptyModulesDirectory = await createFixtureModulesDirectory();
+    running = await startMigratedApi(database.url, emptyModulesDirectory);
+    origin = apiOrigin(running);
+
+    // The owner's existing session survives the restart (sessions live in the database, not the
+    // process), so the same cookie proves ordinary identity behavior is unaffected here too.
+    const listResponse = await fetch(`${origin}/api/v1/grants?userId=${ownerUserId}`, {
+      headers: { cookie: ownerCookie },
+    });
+    const listBody = (await listResponse.json()) as {
+      grants: Array<{ id: string; permission: string }>;
+    };
+    expect(listResponse.status).toBe(200);
+    expect(listBody.grants.some((grant) => grant.id === createdGrantId)).toBe(true);
+
+    const response = await fetch(`${origin}/api/v1/grants`, {
+      body: JSON.stringify({ permission: modulePermission, subjectUserId: ownerUserId }),
+      headers: {
+        "content-type": "application/json",
+        cookie: ownerCookie,
+        origin: dashboardOrigin,
+      },
+      method: "POST",
+    });
+    expect(response.status).toBe(403);
+    expect(await response.json()).toEqual({ code: "AUTHORIZATION_DENIED" });
+
+    const admin = await openAdministrativeClient(database.url);
+    try {
+      const rows = await safeQuery<{ id: string }>(
+        admin,
+        `select id from "ekklesia_core_grant" where id = $1 and revoked_at is null`,
+        [createdGrantId],
+      );
+      expect(rows).toHaveLength(1);
+    } finally {
+      await safeEnd(admin);
+    }
+
+    await running.shutdown();
+    await dropDatabase(database.name);
+  });
+});
diff --git a/apps/api/tests/app.test.ts b/apps/api/tests/app.test.ts
index d2d590c..a7d9e27 100644
--- a/apps/api/tests/app.test.ts
+++ b/apps/api/tests/app.test.ts
@@ -1,5 +1,10 @@
 import { ANONYMOUS_PRINCIPAL, createAuthenticatedPrincipal } from "@ekklesia/contracts";
-import type { AuthorizationEvaluator, AuthorizationStore } from "@ekklesia/core/authorization";
+import {
+  type AuthorizationEvaluator,
+  type AuthorizationStore,
+  CORE_PERMISSION_REGISTRY,
+  type PermissionRegistry,
+} from "@ekklesia/core/authorization";
 import type { EkklesiaIdentityRuntime } from "@ekklesia/core/identity";
 import { IdentityRuntimeError } from "@ekklesia/core/identity";
 import { describe, expect, it, vi } from "vitest";
@@ -17,9 +22,13 @@ function fakeRuntime(overrides: Partial<EkklesiaIdentityRuntime> = {}): Ekklesia
 }
 
 /** A fully stubbed authorization dependency, denying everything by default — tests that need a
- * specific decision or store behavior override the relevant method explicitly. */
+ * specific decision or store behavior override the relevant method explicitly. The registry
+ * defaults to Core's own (containing `core.authorization.grants.manage`, exactly what the
+ * mounted grants routes require) rather than an app-composed one — these tests exercise
+ * `createApiApp`'s wiring, not permission composition itself. */
 function fakeAuthorization(overrides: {
   readonly evaluator?: Partial<AuthorizationEvaluator>;
+  readonly registry?: PermissionRegistry;
   readonly store?: Partial<AuthorizationStore>;
 }): CreateApiAppAuthorizationOptions {
   const evaluator: AuthorizationEvaluator = {
@@ -36,7 +45,7 @@ function fakeAuthorization(overrides: {
     revokePlatformOwner: vi.fn(),
     ...overrides.store,
   };
-  return { evaluator, store };
+  return { evaluator, registry: overrides.registry ?? CORE_PERMISSION_REGISTRY, store };
 }
 
 describe("Ekklesia API Hono application", () => {
@@ -253,6 +262,32 @@ describe("Ekklesia API Hono application", () => {
       expect(evaluatorAuthorize).toHaveBeenCalledTimes(1);
     });
 
+    it("uses exactly the injected registry, never an internal Core-only fallback", async () => {
+      // An evaluator that would allow anything, paired with a registry that does not register
+      // core.authorization.grants.manage at all — if createApiApp silently fell back to some
+      // internal registry containing it, this request would reach the evaluator and succeed
+      // (empty grant list, 200). It must instead fail closed on the registry lookup itself.
+      const evaluatorAuthorize = vi.fn().mockResolvedValue({ allowed: true });
+      const emptyRegistry: PermissionRegistry = {
+        declarations: [],
+        get: () => undefined,
+        has: () => false,
+      };
+      const app = createApiApp({
+        authorization: fakeAuthorization({
+          evaluator: { authorize: evaluatorAuthorize },
+          registry: emptyRegistry,
+        }),
+        runtime: fakeRuntime(),
+        trustedOrigins: [trustedOrigin],
+      });
+
+      const response = await app.request("/api/v1/grants?userId=user_123");
+
+      expect(response.status).toBe(403);
+      expect(evaluatorAuthorize).not.toHaveBeenCalled();
+    });
+
     it("applies the same CORS policy to /api/v1/grants as the rest of /api/v1/*", async () => {
       const app = createApiApp({
         authorization: fakeAuthorization({}),
diff --git a/apps/api/tests/authorization-middleware.test.ts b/apps/api/tests/authorization-middleware.test.ts
index 9c42136..bb4aee2 100644
--- a/apps/api/tests/authorization-middleware.test.ts
+++ b/apps/api/tests/authorization-middleware.test.ts
@@ -151,6 +151,153 @@ describe("requirePermission", () => {
   });
 });
 
+describe("requirePermission with module-owned declarations", () => {
+  // requirePermission has no concept of "Core" versus "module" — it only ever consults whatever
+  // registry it is given. These declarations are shaped exactly like what
+  // composePlatformPermissionRegistry would contribute from a real installed module (an id
+  // namespaced under a module, not `core.*`), proving the shared enforcement mechanism works
+  // identically for module-owned permissions without any module-specific code path.
+  const modulePublicPermission = parsePermissionId("calendar.events.list");
+  const moduleAuthenticatedPermission = parsePermissionId("calendar.events.subscribe");
+  const moduleRestrictedPermission = parsePermissionId("calendar.events.read");
+  const moduleRegistry = createPermissionRegistry([
+    {
+      baseline: "public",
+      description: "List public calendar events.",
+      id: modulePublicPermission,
+    },
+    {
+      baseline: "authenticated",
+      description: "Subscribe to calendar event updates.",
+      id: moduleAuthenticatedPermission,
+    },
+    {
+      baseline: "restricted",
+      description: "Read calendar events.",
+      id: moduleRestrictedPermission,
+    },
+  ]);
+
+  function buildModuleApp(
+    permissionId: typeof modulePublicPermission,
+    options: Partial<AuthorizationMiddlewareOptions> = {},
+  ) {
+    const runtime = options.runtime ?? fakeRuntime();
+    const evaluator = options.evaluator ?? fakeEvaluator();
+    const middlewareOptions: AuthorizationMiddlewareOptions = {
+      evaluator,
+      registry: moduleRegistry,
+      runtime,
+    };
+    const app = new Hono();
+    app.get("/module-protected", requirePermission(middlewareOptions, permissionId), (c) =>
+      c.json({ ok: true }),
+    );
+    return { app, evaluator, runtime };
+  }
+
+  it("resolves a module public permission's trusted baseline and passes it to the evaluator unchanged", async () => {
+    const evaluator = fakeEvaluator();
+    const { app } = buildModuleApp(modulePublicPermission, { evaluator });
+
+    const response = await app.request("/module-protected");
+
+    expect(response.status).toBe(200);
+    expect(evaluator.authorize).toHaveBeenCalledWith(
+      expect.objectContaining({ baseline: "public", permission: modulePublicPermission }),
+    );
+  });
+
+  it("resolves a module authenticated permission's trusted baseline and passes it to the evaluator unchanged", async () => {
+    const evaluator = fakeEvaluator();
+    const { app } = buildModuleApp(moduleAuthenticatedPermission, { evaluator });
+
+    await app.request("/module-protected");
+
+    expect(evaluator.authorize).toHaveBeenCalledWith(
+      expect.objectContaining({
+        baseline: "authenticated",
+        permission: moduleAuthenticatedPermission,
+      }),
+    );
+  });
+
+  it("resolves a module restricted permission's trusted baseline and passes it to the evaluator unchanged", async () => {
+    const evaluator = fakeEvaluator();
+    const { app } = buildModuleApp(moduleRestrictedPermission, { evaluator });
+
+    await app.request("/module-protected");
+
+    expect(evaluator.authorize).toHaveBeenCalledWith(
+      expect.objectContaining({
+        baseline: "restricted",
+        permission: moduleRestrictedPermission,
+      }),
+    );
+  });
+
+  it("allows an anonymous principal for a module permission the evaluator authorizes", async () => {
+    const runtime = fakeRuntime({
+      resolvePrincipal: vi.fn().mockResolvedValue(ANONYMOUS_PRINCIPAL),
+    });
+    const { app } = buildModuleApp(modulePublicPermission, { runtime });
+
+    const response = await app.request("/module-protected");
+
+    expect(response.status).toBe(200);
+  });
+
+  it("allows an authenticated principal for a module permission the evaluator authorizes", async () => {
+    const runtime = fakeRuntime({
+      resolvePrincipal: vi.fn().mockResolvedValue(createAuthenticatedPrincipal("user_1")),
+    });
+    const { app } = buildModuleApp(moduleRestrictedPermission, { runtime });
+
+    const response = await app.request("/module-protected");
+
+    expect(response.status).toBe(200);
+  });
+
+  it("denies a module permission the evaluator does not authorize", async () => {
+    const evaluator = fakeEvaluator({ authorize: vi.fn().mockResolvedValue({ allowed: false }) });
+    const { app } = buildModuleApp(moduleRestrictedPermission, { evaluator });
+
+    const response = await app.request("/module-protected");
+
+    expect(response.status).toBe(403);
+    expect(await response.json()).toEqual({ code: "AUTHORIZATION_DENIED" });
+  });
+
+  it("denies a module permission absent from the registry before ever resolving the principal or consulting the evaluator", async () => {
+    const runtime = fakeRuntime();
+    const evaluator = fakeEvaluator();
+    const absentModulePermission = parsePermissionId("scheduler.jobs.read");
+    const { app } = buildModuleApp(absentModulePermission, { evaluator, runtime });
+
+    const response = await app.request("/module-protected");
+
+    expect(response.status).toBe(403);
+    expect(await response.json()).toEqual({ code: "AUTHORIZATION_DENIED" });
+    expect(runtime.resolvePrincipal).not.toHaveBeenCalled();
+    expect(evaluator.authorize).not.toHaveBeenCalled();
+  });
+
+  it("never lets a client-supplied baseline influence the decision for a module permission", async () => {
+    const evaluator = fakeEvaluator();
+    const { app } = buildModuleApp(moduleRestrictedPermission, { evaluator });
+
+    // A hostile client cannot make a restricted module permission behave as public — there is no
+    // request field the middleware reads for a baseline override.
+    await app.request("/module-protected", {
+      headers: { "x-baseline": "public", "x-permission-baseline": "public" },
+    });
+
+    expect(evaluator.authorize).toHaveBeenCalledWith(
+      expect.objectContaining({ baseline: "restricted" }),
+    );
+  });
+});
+
 describe("getAuthorizedPrincipal", () => {
   it("throws when called on a route requirePermission never protected", async () => {
     const { app } = buildTestApp();
diff --git a/apps/api/tests/grants-routes.test.ts b/apps/api/tests/grants-routes.test.ts
index 0b1b37c..07012a5 100644
--- a/apps/api/tests/grants-routes.test.ts
+++ b/apps/api/tests/grants-routes.test.ts
@@ -25,8 +25,13 @@ const subjectUserId = parseEkklesiaUserId("subject_1");
 // (or, for listing, still return) an unregistered permission id rather than assuming registration.
 const permission = parsePermissionId("calendar.events.read");
 // A permission `CORE_PERMISSION_REGISTRY` actually declares, so tests that must reach
-// `store.createGrant`/`store.revokeGrant` can do so without also depending on module permission
-// registration, which does not exist yet.
+// `store.createGrant`/`store.revokeGrant` can do so. This suite exercises `createGrantsRoutes` in
+// isolation and intentionally injects `CORE_PERMISSION_REGISTRY` directly rather than a composed
+// platform registry — the routes have no notion of "Core" versus "module" permissions, they only
+// ever consult whichever registry they are given. Coverage proving a module-declared permission
+// works through the exact same route code lives in
+// `apps/api/tests/authorization-middleware.test.ts` (unit) and the "installed module permission
+// end-to-end" scenario in `apps/api/tests/api-http.integration.ts` (live).
 const registeredPermission = CORE_AUTHORIZATION_GRANTS_MANAGE;
 const grantId = parseGrantId("11111111-1111-1111-1111-111111111111");
 
diff --git a/apps/api/tests/lifecycle.test.ts b/apps/api/tests/lifecycle.test.ts
index 6e5a3dd..c58cdae 100644
--- a/apps/api/tests/lifecycle.test.ts
+++ b/apps/api/tests/lifecycle.test.ts
@@ -1,4 +1,7 @@
-import { beforeEach, describe, expect, it, vi } from "vitest";
+import { mkdir, mkdtemp, rm, writeFile } from "node:fs/promises";
+import { tmpdir } from "node:os";
+import path from "node:path";
+import { afterEach, beforeEach, describe, expect, it, vi } from "vitest";
 
 const poolConstructor = vi.hoisted(() => vi.fn());
 const betterAuthMock = vi.hoisted(() => vi.fn());
@@ -11,6 +14,13 @@ vi.mock("better-auth", () => ({ betterAuth: betterAuthMock }));
 vi.mock("better-auth/adapters/drizzle", () => ({ drizzleAdapter: drizzleAdapterMock }));
 vi.mock("@hono/node-server", () => ({ serve: serveMock }));
 
+import { parseModuleManifest } from "@ekklesia/contracts";
+import {
+  type DiscoveredModule,
+  ModuleDiscoveryError,
+  ModulePermissionCompositionError,
+} from "@ekklesia/core/modules";
+import { PermissionRegistryError } from "@ekklesia/core/authorization";
 import { createApiRuntimeConfig } from "../src/config.js";
 import { installShutdownSignalHandlers, startApi } from "../src/server.js";
 
@@ -30,6 +40,84 @@ function validConfig() {
   );
 }
 
+const temporaryDirectories: string[] = [];
+
+async function createTemporaryModulesDirectory(): Promise<string> {
+  const directory = await mkdtemp(path.join(tmpdir(), "ekklesia-api-modules-"));
+  temporaryDirectories.push(directory);
+  return directory;
+}
+
+/** Writes a minimal, structurally valid module fixture (package.json + ekklesia.module.json +
+ * server entrypoint) directly to disk — real discovery, not a hand-built double. The server
+ * entrypoint throws immediately if ever imported, so a test that reaches a successful assertion
+ * without that throw firing is direct evidence permission composition never executed it. */
+async function createFixtureModule(
+  modulesDirectory: string,
+  id: string,
+  manifestOverrides: Record<string, unknown> = {},
+): Promise<void> {
+  const moduleDirectory = path.join(modulesDirectory, id);
+  const manifest = {
+    manifestVersion: 3,
+    id,
+    name: `${id} module`,
+    version: "0.1.0",
+    compatibility: { platform: "^0.1.0", moduleSdk: "^0.1.0" },
+    permissions: [],
+    entrypoints: { server: "./dist/server/index.js" },
+    ...manifestOverrides,
+  };
+  await mkdir(path.join(moduleDirectory, "dist/server"), { recursive: true });
+  await Promise.all([
+    writeFile(
+      path.join(moduleDirectory, "package.json"),
+      JSON.stringify({
+        ekklesia: { layer: "module" },
+        name: `@test/module-${id}`,
+        type: "module",
+        version: "0.1.0",
+      }),
+    ),
+    writeFile(path.join(moduleDirectory, "ekklesia.module.json"), JSON.stringify(manifest)),
+    writeFile(
+      path.join(moduleDirectory, "dist/server/index.js"),
+      'throw new Error("permission composition must never execute a module server entrypoint");',
+    ),
+  ]);
+}
+
+/** A minimal, in-memory `DiscoveredModule` double built from a real, validated manifest — used
+ * only to reach `composePlatformPermissionRegistry`'s own duplicate-detection failure, which is
+ * unreachable through real filesystem discovery (module ids are unique and every permission id is
+ * namespaced by its own module's id, so no two genuinely discovered modules can ever collide). */
+function fakeDiscoveredModule(
+  id: string,
+  permissions: ReadonlyArray<{
+    id: string;
+    baseline: "public" | "authenticated" | "restricted";
+    description: string;
+  }>,
+): DiscoveredModule {
+  const manifest = parseModuleManifest({
+    capabilities: [],
+    compatibility: { moduleSdk: "^0.1.0", platform: "^0.1.0" },
+    entrypoints: { server: "./dist/does-not-exist/index.js" },
+    id,
+    manifestVersion: 3,
+    name: `${id} module`,
+    permissions,
+    version: "0.1.0",
+  });
+  return {
+    directory: `/modules/${id}`,
+    entrypoints: { server: `/modules/${id}/dist/does-not-exist/index.js` },
+    manifest,
+    manifestPath: `/modules/${id}/ekklesia.module.json`,
+    packageName: `@test/${id}`,
+  };
+}
+
 /** A minimal fake `ServerType`: enough of Node's `http.Server` shape for `bindServer`/`shutdown`
  * to drive without a real socket. `close` and `once` are spies so tests can assert call counts;
  * `triggerError` simulates the server's own `'error'` event (e.g. a bind failure). */
@@ -285,3 +373,148 @@ describe("Ekklesia API shutdown lifecycle", () => {
     restore();
   });
 });
+
+describe("Ekklesia API module permission composition at startup", () => {
+  afterEach(async () => {
+    await Promise.all(
+      temporaryDirectories.splice(0).map((directory) => rm(directory, { recursive: true })),
+    );
+  });
+
+  beforeEach(() => {
+    poolConstructor.mockReset();
+    betterAuthMock.mockReset();
+    drizzleAdapterMock.mockClear();
+    serveMock.mockReset();
+  });
+
+  it("discovers and composes permissions before constructing the database pool, Better Auth, or the HTTP server", async () => {
+    const modulesDirectory = await createTemporaryModulesDirectory();
+    // An incompatible platform version fails discovery itself (MODULE_PLATFORM_INCOMPATIBLE),
+    // before permission composition — and, since composition runs before any pool exists, before
+    // any pool, Better Auth instance, or HTTP server could ever be constructed.
+    await createFixtureModule(modulesDirectory, "incompatible", {
+      compatibility: { moduleSdk: "^0.1.0", platform: ">=99.0.0" },
+    });
+
+    await expect(
+      startApi(validConfig(), { modulesDirectoryOverride: modulesDirectory }),
+    ).rejects.toBeInstanceOf(ModuleDiscoveryError);
+
+    expect(poolConstructor).not.toHaveBeenCalled();
+    expect(betterAuthMock).not.toHaveBeenCalled();
+    expect(serveMock).not.toHaveBeenCalled();
+  });
+
+  it("fails startup when a module claims the reserved core.* permission namespace, creating no pool, Better Auth instance, or server", async () => {
+    const modulesDirectory = await createTemporaryModulesDirectory();
+    // Only a module whose own id is literally "core" can produce a "core."-prefixed permission
+    // that still passes the manifest's own module-namespace check.
+    await createFixtureModule(modulesDirectory, "core", {
+      permissions: [
+        { baseline: "restricted", description: "Not really Core's.", id: "core.rogue.permission" },
+      ],
+    });
+
+    let captured: unknown;
+    try {
+      await startApi(validConfig(), { modulesDirectoryOverride: modulesDirectory });
+    } catch (error) {
+      captured = error;
+    }
+
+    expect(captured).toBeInstanceOf(ModulePermissionCompositionError);
+    expect(captured).toMatchObject({ code: "MODULE_PERMISSION_RESERVED_NAMESPACE" });
+    expect(poolConstructor).not.toHaveBeenCalled();
+    expect(betterAuthMock).not.toHaveBeenCalled();
+    expect(serveMock).not.toHaveBeenCalled();
+  });
+
+  it("fails startup on a duplicate permission declaration, creating no pool, Better Auth instance, or server", async () => {
+    const first = fakeDiscoveredModule("calendar", [
+      { baseline: "restricted", description: "Read events.", id: "calendar.events.read" },
+    ]);
+    const second = fakeDiscoveredModule("calendar", [
+      { baseline: "public", description: "Also read events.", id: "calendar.events.read" },
+    ]);
+
+    let captured: unknown;
+    try {
+      await startApi(validConfig(), { discoveredModulesOverride: [first, second] });
+    } catch (error) {
+      captured = error;
+    }
+
+    expect(captured).toBeInstanceOf(PermissionRegistryError);
+    expect(captured).toMatchObject({ code: "PERMISSION_REGISTRY_DUPLICATE_DECLARATION" });
+    expect(poolConstructor).not.toHaveBeenCalled();
+    expect(betterAuthMock).not.toHaveBeenCalled();
+    expect(serveMock).not.toHaveBeenCalled();
+  });
+
+  it("keeps a reserved-namespace failure and a generic duplicate-declaration failure distinguishable", async () => {
+    const modulesDirectory = await createTemporaryModulesDirectory();
+    await createFixtureModule(modulesDirectory, "core", {
+      permissions: [
+        {
+          baseline: "restricted",
+          description: "Attempting to shadow Core's own permission.",
+          id: "core.authorization.grants.manage",
+        },
+      ],
+    });
+
+    let reservedNamespaceError: unknown;
+    try {
+      await startApi(validConfig(), { modulesDirectoryOverride: modulesDirectory });
+    } catch (error) {
+      reservedNamespaceError = error;
+    }
+
+    const duplicateModule = fakeDiscoveredModule("calendar", [
+      { baseline: "restricted", description: "Read events.", id: "calendar.events.read" },
+    ]);
+    const duplicateModuleAgain = fakeDiscoveredModule("calendar", [
+      { baseline: "restricted", description: "Read events, again.", id: "calendar.events.read" },
+    ]);
+    let duplicateError: unknown;
+    try {
+      await startApi(validConfig(), {
+        discoveredModulesOverride: [duplicateModule, duplicateModuleAgain],
+      });
+    } catch (error) {
+      duplicateError = error;
+    }
+
+    // Even though the first scenario's module attempted to exactly collide with an existing
+    // Core-owned permission id, it is reported as a reserved-namespace violation, never as a
+    // generic duplicate — the two failure families never overlap.
+    expect(reservedNamespaceError).toBeInstanceOf(ModulePermissionCompositionError);
+    expect(reservedNamespaceError).not.toBeInstanceOf(PermissionRegistryError);
+    expect(duplicateError).toBeInstanceOf(PermissionRegistryError);
+    expect(duplicateError).not.toBeInstanceOf(ModulePermissionCompositionError);
+  });
+
+  it("still constructs exactly one pool and succeeds when zero modules are installed", async () => {
+    const modulesDirectory = await createTemporaryModulesDirectory();
+    const poolEnd = vi.fn().mockResolvedValue(undefined);
+    poolConstructor.mockImplementation(function fakePool() {
+      return { end: poolEnd, on: vi.fn() };
+    });
+    betterAuthMock.mockReturnValue({ api: { getSession: vi.fn() }, handler: vi.fn() });
+    mockSuccessfulServe();
+
+    const running = await startApi(validConfig(), { modulesDirectoryOverride: modulesDirectory });
+
+    expect(poolConstructor).toHaveBeenCalledTimes(1);
+    await running.shutdown();
+    expect(poolEnd).toHaveBeenCalledTimes(1);
+  });
+
+  it("does not perform module discovery merely by importing the platform-permissions module", async () => {
+    const module = await import("../src/platform-permissions.js");
+
+    expect(typeof module.composeStartupPermissionRegistry).toBe("function");
+    expect(poolConstructor).not.toHaveBeenCalled();
+  });
+});
diff --git a/core/src/authorization/permissions.ts b/core/src/authorization/permissions.ts
index 66e52ce..e81e56f 100644
--- a/core/src/authorization/permissions.ts
+++ b/core/src/authorization/permissions.ts
@@ -1,25 +1,18 @@
 import {
-  type PermissionBaseline,
+  type PermissionDeclaration,
   type PermissionId,
   parsePermissionId,
   safeParsePermissionId,
 } from "@ekklesia/contracts";
 
 /**
- * A registered permission declaration: the trusted pairing of a permission identifier with its
- * one baseline. This is what distinguishes a *registered* permission from a merely well-formed
- * `PermissionId` string — the evaluator itself (`createAuthorizationEvaluator`) still takes an
- * explicit, caller-supplied baseline per pass 1's design; the registry is what a trusted caller
- * (Core's own HTTP enforcement, eventually a module registration mechanism) consults so that
- * baseline is never client-supplied or guessed.
+ * Core's registry consumes the one shared `PermissionDeclaration` shape
+ * (`packages/contracts/src/authorization.ts`) rather than declaring its own interface — a module
+ * manifest's `permissions` array parses into the exact same type, so a declaration a module ships
+ * and a declaration Core registers are provably interchangeable. Re-exported here so existing
+ * `@ekklesia/core/authorization` consumers keep importing it from this package.
  */
-export interface PermissionDeclaration {
-  readonly id: PermissionId;
-  readonly baseline: PermissionBaseline;
-  /** Human-readable, for operators and audit review only — never interpreted by the evaluator or
-   * by enforcement. */
-  readonly description: string;
-}
+export type { PermissionDeclaration, PermissionDeclarationInput } from "@ekklesia/contracts";
 
 export interface PermissionRegistry {
   readonly declarations: readonly PermissionDeclaration[];
@@ -53,9 +46,10 @@ export class PermissionRegistryError extends Error {
  * Builds an immutable permission registry from a fixed list of declarations, rejecting any
  * duplicate permission id outright at composition time. Deliberately the smallest mechanism that
  * distinguishes "registered" from "well-formed": no module discovery, no dynamic registration, no
- * network or filesystem I/O — a future module registration mechanism can merge its own
- * declarations into a combined list and still call this same constructor, inheriting duplicate
- * detection for free instead of reimplementing it.
+ * network or filesystem I/O. The module integration layer's composer
+ * (`composePlatformPermissionRegistry`, `core/src/modules/permissions.ts`) merges Core's own
+ * declarations with every discovered module's into one combined list and calls this same
+ * constructor, inheriting duplicate detection for free rather than reimplementing it.
  */
 export function createPermissionRegistry(
   declarations: readonly PermissionDeclaration[],
@@ -89,7 +83,14 @@ export const CORE_AUTHORIZATION_GRANTS_MANAGE: PermissionId = parsePermissionId(
   "core.authorization.grants.manage",
 );
 
-const CORE_PERMISSION_DECLARATIONS: readonly PermissionDeclaration[] = Object.freeze([
+/**
+ * Core's own declarations, exported in deeply immutable form so a trusted composer (the module
+ * integration layer, `core/src/modules/permissions.ts`) can build one combined registry from
+ * exactly these plus every discovered module's declarations, in a fixed, deterministic order,
+ * without reconstructing or guessing Core's own list. `CORE_PERMISSION_REGISTRY` below remains the
+ * registry Core-only code and existing tests use directly.
+ */
+export const CORE_PERMISSION_DECLARATIONS: readonly PermissionDeclaration[] = Object.freeze([
   Object.freeze({
     baseline: "restricted" as const,
     description: "Create, list, and revoke another user's explicit authorization grants.",
@@ -98,11 +99,12 @@ const CORE_PERMISSION_DECLARATIONS: readonly PermissionDeclaration[] = Object.fr
 ]);
 
 /**
- * Core's own permission registry — the trusted source of truth `apps/api`'s authorization
- * enforcement consults so an HTTP route can never supply its own baseline. Contains exactly the
- * one Core-owned permission this slice needs; a future module registration mechanism is
- * deliberately out of scope here and will compose its own declarations through
- * `createPermissionRegistry` rather than through this constant.
+ * Core's own permission registry, containing exactly the Core-owned permissions declared above.
+ * Production `apps/api` does not consult this constant directly: it enforces requests against the
+ * composed platform registry (`composePlatformPermissionRegistry`,
+ * `core/src/modules/permissions.ts`), which merges these same declarations with every installed
+ * module's. `CORE_PERMISSION_REGISTRY` remains for Core-only callers and tests that need Core's
+ * permissions in isolation, without composing in any module.
  */
 export const CORE_PERMISSION_REGISTRY: PermissionRegistry = createPermissionRegistry(
   CORE_PERMISSION_DECLARATIONS,
diff --git a/core/src/cli/validate-modules.ts b/core/src/cli/validate-modules.ts
index 5e7b7ba..71b3f70 100644
--- a/core/src/cli/validate-modules.ts
+++ b/core/src/cli/validate-modules.ts
@@ -1,8 +1,11 @@
 import { readFile } from "node:fs/promises";
 import path from "node:path";
+import { PermissionRegistryError } from "../authorization/index.js";
 import {
+  composePlatformPermissionRegistry,
   discoverModuleSchemas,
   ModuleDiscoveryError,
+  ModulePermissionCompositionError,
   ModuleSchemaDiscoveryError,
 } from "../modules/index.js";
 
@@ -31,6 +34,12 @@ async function main(): Promise<void> {
   });
   const modules = schemaAggregate.discoveredModules;
 
+  // Reuses the modules discoverModuleSchemas already discovered — never rediscovers modules
+  // solely for permission composition. A manifest that passes structural validation but claims
+  // the reserved core.* namespace, or duplicates another declaration, fails validation here
+  // before this module is considered installable.
+  composePlatformPermissionRegistry(modules);
+
   if (modules.length === 0) {
     console.log("Module validation passed (0 modules installed).");
     return;
@@ -57,6 +66,10 @@ main().catch((error: unknown) => {
       const issuePath = issue.path ? ` (${issue.path})` : "";
       console.error(`[${issue.code}] ${issue.moduleId}${issuePath}: ${issue.message}`);
     }
+  } else if (error instanceof ModulePermissionCompositionError) {
+    console.error(`[${error.code}] ${error.message}`);
+  } else if (error instanceof PermissionRegistryError) {
+    console.error(`[${error.code}] ${error.message}`);
   } else {
     console.error(error instanceof Error ? error.message : String(error));
   }
diff --git a/core/src/diagnostics/doctor-report.ts b/core/src/diagnostics/doctor-report.ts
index e9a5267..bd4003c 100644
--- a/core/src/diagnostics/doctor-report.ts
+++ b/core/src/diagnostics/doctor-report.ts
@@ -11,7 +11,7 @@ function section(title: string, checkIds: readonly DoctorCheckId[]): DoctorSecti
 
 const sections: readonly DoctorSection[] = Object.freeze([
   section("Platform", ["platform.runtime"]),
-  section("Modules", ["modules.discovery", "modules.dependencies"]),
+  section("Modules", ["modules.discovery", "modules.dependencies", "modules.permissions"]),
   section("Data definitions", ["schemas.discovery", "migrations.discovery"]),
   section("PostgreSQL", ["postgresql.connection", "postgresql.plan"]),
   section("Identity", ["identity.configuration"]),
diff --git a/core/src/diagnostics/doctor.ts b/core/src/diagnostics/doctor.ts
index 53498ac..88934a5 100644
--- a/core/src/diagnostics/doctor.ts
+++ b/core/src/diagnostics/doctor.ts
@@ -1,6 +1,12 @@
 import path from "node:path";
-import { isVersionCompatible, moduleIdSchema, semanticVersionSchema } from "@ekklesia/contracts";
+import {
+  isVersionCompatible,
+  moduleIdSchema,
+  permissionIdSchema,
+  semanticVersionSchema,
+} from "@ekklesia/contracts";
 import { getTableName } from "drizzle-orm";
+import { type PermissionRegistry, PermissionRegistryError } from "../authorization/index.js";
 import {
   type CreateIdentityConfigOptions,
   createIdentityConfig,
@@ -18,9 +24,11 @@ import {
   planModuleMigrations,
 } from "../migrations/index.js";
 import {
+  composePlatformPermissionRegistry,
   type DiscoveredModule,
   discoverModuleSchemas,
   discoverModules,
+  ModulePermissionCompositionError,
   type ModuleDiscoveryOptions,
   type ModuleMigrationAggregate,
   type ModuleSchemaAggregate,
@@ -30,6 +38,7 @@ export type DoctorCheckId =
   | "platform.runtime"
   | "modules.discovery"
   | "modules.dependencies"
+  | "modules.permissions"
   | "schemas.discovery"
   | "migrations.discovery"
   | "postgresql.connection"
@@ -68,6 +77,9 @@ export interface DoctorOptions {
 }
 
 export interface DoctorDependencies {
+  /** Synchronous, matching the real implementation: pure in-memory composition over an
+   * already-discovered module list, never re-discovering modules itself. */
+  readonly composePermissions: (modules: readonly DiscoveredModule[]) => PermissionRegistry;
   readonly createMigrationDatabase: (
     connectionString: string,
     options: PostgresMigrationDatabaseOptions,
@@ -101,6 +113,7 @@ async function discoverAllMigrationsForRuntime(
 }
 
 const productionDependencies: DoctorDependencies = Object.freeze({
+  composePermissions: composePlatformPermissionRegistry,
   createMigrationDatabase: createPostgresMigrationDatabase,
   discoverMigrations: discoverAllMigrationsForRuntime,
   discoverModules,
@@ -163,6 +176,57 @@ function safeMigrationReference(identity: string, checksum: string): string {
   return `${identity} · ${checksum.slice(0, 12)}`;
 }
 
+function safeModuleId(value: string): string {
+  if (!moduleIdSchema.safeParse(value).success) {
+    throw new TypeError("Unsafe module diagnostic value.");
+  }
+  return value;
+}
+
+function safePermissionId(value: string): string {
+  if (!permissionIdSchema.safeParse(value).success) {
+    throw new TypeError("Unsafe permission diagnostic value.");
+  }
+  return value;
+}
+
+/** Distinguishes the two composition-error families rather than describing a reserved `core.*`
+ * declaration as a generic duplicate. Never includes a module directory, manifest path, caught
+ * cause, or stack — only the module id and permission id the structured errors already carry,
+ * each re-validated before being interpolated (defense in depth against a future drift bug,
+ * matching `safeModuleDetails`/`safeMigrationReference` elsewhere in this file); if that
+ * re-validation itself fails, this falls back to a fixed generic detail rather than throwing. */
+function permissionCompositionFailure(error: unknown): DoctorCheck {
+  if (error instanceof ModulePermissionCompositionError) {
+    try {
+      const moduleId = safeModuleId(error.moduleId);
+      const permissionId = safePermissionId(error.permissionId);
+      return freezeCheck("modules.permissions", "failure", "Permission composition failed", [
+        `Module "${moduleId}" declared permission "${permissionId}" in the reserved core.* namespace.`,
+      ]);
+    } catch {
+      return freezeCheck("modules.permissions", "failure", "Permission composition failed", [
+        "A module declared a permission in the reserved core.* namespace.",
+      ]);
+    }
+  }
+  if (error instanceof PermissionRegistryError) {
+    try {
+      const permissionId = safePermissionId(error.permissionId);
+      return freezeCheck("modules.permissions", "failure", "Permission composition failed", [
+        `Permission "${permissionId}" is declared more than once.`,
+      ]);
+    } catch {
+      return freezeCheck("modules.permissions", "failure", "Permission composition failed", [
+        "A permission is declared more than once.",
+      ]);
+    }
+  }
+  return freezeCheck("modules.permissions", "failure", "Permission composition failed", [
+    "The installed permission declarations could not be composed safely.",
+  ]);
+}
+
 function moduleDiscoveryFailure(error: unknown): DoctorCheck {
   const details =
     error instanceof Error && error.name === "ModuleDiscoveryError"
@@ -386,6 +450,33 @@ export async function diagnoseEkklesia(
     ]);
   }
 
+  // Reuses the already-discovered module list above — never re-discovers modules solely for
+  // permission composition. Independent of the schema/migration checks below: a permission
+  // composition failure here (a reserved-namespace or duplicate declaration) does not skip those,
+  // since neither depends on the permission registry, only on discovery having succeeded.
+  let permissionsCheck: DoctorCheck;
+  if (!discoveredModules) {
+    permissionsCheck = freezeCheck(
+      "modules.permissions",
+      "skipped",
+      "Permission composition skipped",
+      ["Complete module discovery is required."],
+    );
+  } else {
+    try {
+      const registry = dependencies.composePermissions(discoveredModules);
+      const count = registry.declarations.length;
+      permissionsCheck = freezeCheck(
+        "modules.permissions",
+        "pass",
+        "Permission registry composed",
+        [`${count} ${plural(count, "permission")} registered`],
+      );
+    } catch (error) {
+      permissionsCheck = permissionCompositionFailure(error);
+    }
+  }
+
   let schemaCheck: DoctorCheck;
   let migrationCheck: DoctorCheck;
   let migrations: ModuleMigrationAggregate | undefined;
@@ -466,6 +557,7 @@ export async function diagnoseEkklesia(
     runtimeCheck,
     moduleCheck,
     dependencyCheck,
+    permissionsCheck,
     schemaCheck,
     migrationCheck,
     postgresConnectionCheck,
diff --git a/core/src/modules/discovery.ts b/core/src/modules/discovery.ts
index ca497ed..7c133bf 100644
--- a/core/src/modules/discovery.ts
+++ b/core/src/modules/discovery.ts
@@ -47,13 +47,19 @@ export interface ResolvedModuleEntrypoints {
   readonly jobs?: string;
 }
 
-type DeepReadonly<T> = T extends (...arguments_: never[]) => unknown
+// Branded primitives (`PermissionId`, `ModuleId`, ...) are `string & {brand}` intersection types,
+// which structurally satisfy `T extends object` — without this check first, the mapped-type
+// branch below would incorrectly expand a branded string into `String.prototype`'s member shape
+// and lose the brand entirely, rather than leaving an already-immutable primitive untouched.
+type DeepReadonly<T> = T extends string | number | boolean | bigint | symbol | null | undefined
   ? T
-  : T extends readonly (infer Item)[]
-    ? readonly DeepReadonly<Item>[]
-    : T extends object
-      ? { readonly [Key in keyof T]: DeepReadonly<T[Key]> }
-      : T;
+  : T extends (...arguments_: never[]) => unknown
+    ? T
+    : T extends readonly (infer Item)[]
+      ? readonly DeepReadonly<Item>[]
+      : T extends object
+        ? { readonly [Key in keyof T]: DeepReadonly<T[Key]> }
+        : T;
 
 export type ReadonlyModuleManifest = DeepReadonly<ModuleManifest>;
 
diff --git a/core/src/modules/index.ts b/core/src/modules/index.ts
index 085b2e4..b324e7e 100644
--- a/core/src/modules/index.ts
+++ b/core/src/modules/index.ts
@@ -1,4 +1,5 @@
 export * from "./discovery.js";
 export * from "./migration-discovery.js";
+export * from "./permissions.js";
 export * from "./runtime.js";
 export * from "./schema-discovery.js";
diff --git a/core/src/modules/permissions.ts b/core/src/modules/permissions.ts
new file mode 100644
index 0000000..dc82a93
--- /dev/null
+++ b/core/src/modules/permissions.ts
@@ -0,0 +1,82 @@
+import {
+  CORE_PERMISSION_DECLARATIONS,
+  createPermissionRegistry,
+  type PermissionDeclaration,
+  type PermissionRegistry,
+} from "../authorization/index.js";
+import type { DiscoveredModule } from "./discovery.js";
+
+const CORE_PERMISSION_NAMESPACE_PREFIX = "core.";
+
+export type ModulePermissionCompositionErrorCode = "MODULE_PERMISSION_RESERVED_NAMESPACE";
+
+/** Thrown only at composition time, never per-request. Carries the offending module and
+ * permission id deliberately — like `PermissionRegistryError`, this is a startup-time
+ * configuration error, not a runtime response, and neither value is sensitive. Never carries a
+ * module directory, manifest path, caught error, or stack detail. */
+export class ModulePermissionCompositionError extends Error {
+  readonly code: ModulePermissionCompositionErrorCode;
+  readonly moduleId: string;
+  readonly permissionId: string;
+
+  constructor(moduleId: string, permissionId: string) {
+    super(
+      `Module "${moduleId}" cannot declare permission "${permissionId}": the "core." namespace is reserved for Core's own permissions.`,
+    );
+    this.name = "ModulePermissionCompositionError";
+    this.code = "MODULE_PERMISSION_RESERVED_NAMESPACE";
+    this.moduleId = moduleId;
+    this.permissionId = permissionId;
+    this.stack = `${this.name}: ${this.message}`;
+    Object.freeze(this);
+  }
+}
+
+/**
+ * Core's module-integration-layer composition helper: builds the one platform permission
+ * registry from Core's own declarations (`CORE_PERMISSION_DECLARATIONS`) plus every already
+ * discovered module's manifest-declared permissions. Deliberately lives in `core/src/modules`,
+ * not `core/src/authorization` — the dependency runs one way only, this layer depends on
+ * authorization's `createPermissionRegistry`, and authorization never imports module discovery
+ * or runtime. This is pure, synchronous, in-memory composition over already-discovered,
+ * already-validated data: no filesystem, environment, database, network, or module-code
+ * execution happens here, and the caller (not this function) is responsible for having already
+ * produced a complete, successfully discovered module list (for example via `discoverModules`).
+ *
+ * Declaration order is deterministic and meaningful for duplicate detection: Core's own
+ * declarations first, then each module in the exact order it appears in `modules`, and within a
+ * module its declarations in manifest-array order. Duplicate exact permission ids fail via the
+ * same `PermissionRegistryError` `createPermissionRegistry` already throws for any duplicate;
+ * this function does not reimplement or duplicate that detection.
+ *
+ * `core.*` is reserved for Core's own permissions. A discovered module contributing any
+ * `core.`-prefixed permission id — regardless of that module's own id, and regardless of whether
+ * it happens to exactly collide with an existing declaration — fails composition with
+ * `ModulePermissionCompositionError` rather than silently registering or silently colliding. This
+ * check runs before a declaration is ever added to the list `createPermissionRegistry` sees, so it
+ * is also what actually catches a module attempting to collide with a Core-owned id specifically:
+ * every Core-owned id is itself `core.`-prefixed, module ids may never contain a dot (so only a
+ * module whose own id is literally `"core"` can produce a `core.`-prefixed permission at all), and
+ * a genuine module-vs-module collision is structurally unreachable through real discovery — module
+ * ids are unique (discovery itself rejects a duplicate) and every permission id is namespaced by
+ * its own module's id. `PermissionRegistryError` therefore never fires for a real, validly
+ * discovered installed-module set; `createPermissionRegistry`'s duplicate rejection remains here as
+ * a defensive, unconditional guarantee regardless, and is exercised directly with deliberately
+ * conflicting test data rather than through a scenario real discovery could ever produce.
+ */
+export function composePlatformPermissionRegistry(
+  modules: readonly DiscoveredModule[],
+): PermissionRegistry {
+  const declarations: PermissionDeclaration[] = [...CORE_PERMISSION_DECLARATIONS];
+
+  for (const module of modules) {
+    for (const declaration of module.manifest.permissions) {
+      if (declaration.id.startsWith(CORE_PERMISSION_NAMESPACE_PREFIX)) {
+        throw new ModulePermissionCompositionError(module.manifest.id, declaration.id);
+      }
+      declarations.push(declaration);
+    }
+  }
+
+  return createPermissionRegistry(declarations);
+}
diff --git a/core/tests/authorization-permissions.test.ts b/core/tests/authorization-permissions.test.ts
index fb4574a..d6574aa 100644
--- a/core/tests/authorization-permissions.test.ts
+++ b/core/tests/authorization-permissions.test.ts
@@ -2,6 +2,7 @@ import { parsePermissionId } from "@ekklesia/contracts";
 import { describe, expect, it } from "vitest";
 import {
   CORE_AUTHORIZATION_GRANTS_MANAGE,
+  CORE_PERMISSION_DECLARATIONS,
   CORE_PERMISSION_REGISTRY,
   createPermissionRegistry,
   PermissionRegistryError,
@@ -92,3 +93,21 @@ describe("CORE_PERMISSION_REGISTRY", () => {
     expect(CORE_PERMISSION_REGISTRY.declarations).toHaveLength(1);
   });
 });
+
+describe("CORE_PERMISSION_DECLARATIONS", () => {
+  it("is Core's own declaration list, exported separately from the registry in deeply immutable form", () => {
+    expect(CORE_PERMISSION_DECLARATIONS).toEqual([
+      {
+        baseline: "restricted",
+        description: "Create, list, and revoke another user's explicit authorization grants.",
+        id: CORE_AUTHORIZATION_GRANTS_MANAGE,
+      },
+    ]);
+    expect(Object.isFrozen(CORE_PERMISSION_DECLARATIONS)).toBe(true);
+    expect(Object.isFrozen(CORE_PERMISSION_DECLARATIONS[0])).toBe(true);
+  });
+
+  it("matches exactly what CORE_PERMISSION_REGISTRY exposes as its own declarations, so a composer building on top of it never diverges from what Core itself registers", () => {
+    expect(CORE_PERMISSION_DECLARATIONS).toEqual(CORE_PERMISSION_REGISTRY.declarations);
+  });
+});
diff --git a/core/tests/core-migration-discovery.test.ts b/core/tests/core-migration-discovery.test.ts
index 6633404..76b8a52 100644
--- a/core/tests/core-migration-discovery.test.ts
+++ b/core/tests/core-migration-discovery.test.ts
@@ -91,7 +91,7 @@ async function createModule(
     writeFile(
       path.join(moduleDirectory, "ekklesia.module.json"),
       JSON.stringify({
-        manifestVersion: 2,
+        manifestVersion: 3,
         id,
         name: `${id} module`,
         version: "0.1.0",
diff --git a/core/tests/doctor-report.test.ts b/core/tests/doctor-report.test.ts
index ac7349d..7a11f49 100644
--- a/core/tests/doctor-report.test.ts
+++ b/core/tests/doctor-report.test.ts
@@ -6,6 +6,7 @@ import {
 } from "../src/diagnostics/index.js";
 import type { MigrationDatabase } from "../src/migrations/index.js";
 import type { ModuleMigrationAggregate, ModuleSchemaAggregate } from "../src/modules/index.js";
+import { composePlatformPermissionRegistry } from "../src/modules/permissions.js";
 
 const originalNoColor = process.env.NO_COLOR;
 
@@ -36,6 +37,7 @@ function zeroMigrationAggregate(): ModuleMigrationAggregate {
 
 function noConfigurationDependencies(): DoctorDependencies {
   return {
+    composePermissions: composePlatformPermissionRegistry,
     createMigrationDatabase: () => {
       throw new Error("database must not be constructed");
     },
@@ -75,6 +77,8 @@ Modules
     0 modules installed
   ✓ Dependency graph valid
     Dependency order is empty.
+  ✓ Permission registry composed
+    1 permission registered
 
 Data definitions
   ✓ Schemas validated
diff --git a/core/tests/doctor.test.ts b/core/tests/doctor.test.ts
index 9854d2c..1a76ac5 100644
--- a/core/tests/doctor.test.ts
+++ b/core/tests/doctor.test.ts
@@ -1,5 +1,6 @@
-import type { ModuleManifest } from "@ekklesia/contracts";
+import { parsePermissionId, type ModuleManifest } from "@ekklesia/contracts";
 import { describe, expect, it, vi } from "vitest";
+import { createPermissionRegistry, PermissionRegistryError } from "../src/authorization/index.js";
 import {
   type DoctorCheck,
   type DoctorDependencies,
@@ -21,6 +22,7 @@ import type {
   ModuleMigrationAggregate,
   ModuleSchemaAggregate,
 } from "../src/modules/index.js";
+import { ModulePermissionCompositionError } from "../src/modules/permissions.js";
 
 const checksumA = "a".repeat(64);
 const checksumB = "b".repeat(64);
@@ -39,7 +41,7 @@ function discoveredModule(id: string, version = "0.1.0"): DiscoveredModule {
       },
       entrypoints: { server: "./dist/server.js" },
       id,
-      manifestVersion: 2,
+      manifestVersion: 3,
       name: `${id} module`,
       permissions: [],
       storage: { buckets: [] },
@@ -125,6 +127,11 @@ function plan(entries: MigrationPlanResult["entries"] = []): MigrationPlanResult
 
 function dependencies(overrides: Partial<DoctorDependencies> = {}): DoctorDependencies {
   return {
+    composePermissions: vi.fn((modules: readonly DiscoveredModule[]) =>
+      createPermissionRegistry(
+        modules.flatMap((discovered) => [...discovered.manifest.permissions]),
+      ),
+    ),
     createMigrationDatabase: vi.fn(() => emptyDatabase()),
     discoverMigrations: vi.fn(async () => migrationAggregate()),
     discoverModules: vi.fn(async () => Object.freeze([])),
@@ -196,6 +203,7 @@ describe("Ekklesia Doctor diagnostics", () => {
       "platform.runtime",
       "modules.discovery",
       "modules.dependencies",
+      "modules.permissions",
       "schemas.discovery",
       "migrations.discovery",
       "postgresql.connection",
@@ -205,6 +213,10 @@ describe("Ekklesia Doctor diagnostics", () => {
     ]);
     expect(check(report, "modules.discovery").details).toEqual(["people 0.1.0", "calendar 0.2.0"]);
     expect(check(report, "modules.dependencies").details).toEqual(["Order: people → calendar"]);
+    expect(check(report, "modules.permissions")).toMatchObject({
+      details: ["0 permissions registered"],
+      status: "pass",
+    });
     expect(check(report, "migrations.discovery").details).toEqual([
       "1 migration · checksums valid",
       "people:0000_initial · aaaaaaaaaaaa",
@@ -369,6 +381,7 @@ describe("Ekklesia Doctor diagnostics", () => {
     expect(check(report, "modules.discovery").status).toBe("failure");
     for (const id of [
       "modules.dependencies",
+      "modules.permissions",
       "schemas.discovery",
       "migrations.discovery",
       "postgresql.connection",
@@ -385,6 +398,114 @@ describe("Ekklesia Doctor diagnostics", () => {
     expect(JSON.stringify(report)).not.toContain("password");
   });
 
+  it("passes permission composition with a safe declaration count when modules are installed", async () => {
+    const calendar = discoveredModule("calendar");
+    const report = await diagnoseEkklesia(
+      options(),
+      dependencies({
+        composePermissions: vi.fn(() =>
+          createPermissionRegistry([
+            {
+              baseline: "restricted",
+              description: "Read calendar events.",
+              id: parsePermissionId("calendar.events.read"),
+            },
+          ]),
+        ),
+        discoverModules: vi.fn(async () => Object.freeze([calendar])),
+      }),
+    );
+
+    expect(check(report, "modules.permissions")).toMatchObject({
+      details: ["1 permission registered"],
+      status: "pass",
+    });
+  });
+
+  it("reuses the already-discovered module list rather than rediscovering modules for permission composition", async () => {
+    const calendar = discoveredModule("calendar");
+    const composePermissions = vi.fn(() => createPermissionRegistry([]));
+    const discoverModules = vi.fn(async () => Object.freeze([calendar]));
+    await diagnoseEkklesia(options(), dependencies({ composePermissions, discoverModules }));
+
+    expect(discoverModules).toHaveBeenCalledOnce();
+    expect(composePermissions).toHaveBeenCalledOnce();
+    expect(composePermissions).toHaveBeenCalledWith([calendar]);
+  });
+
+  it("fails permission composition when a module claims the reserved core.* namespace, distinctly from a generic duplicate", async () => {
+    const report = await diagnoseEkklesia(
+      options(),
+      dependencies({
+        composePermissions: vi.fn(() => {
+          throw new ModulePermissionCompositionError("core", "core.rogue.permission");
+        }),
+      }),
+    );
+
+    expect(check(report, "modules.permissions")).toMatchObject({
+      details: [
+        'Module "core" declared permission "core.rogue.permission" in the reserved core.* namespace.',
+      ],
+      status: "failure",
+    });
+    expect(report.result).toBe("not-ready");
+  });
+
+  it("fails permission composition on a generic duplicate declaration, distinctly from a reserved-namespace violation", async () => {
+    const report = await diagnoseEkklesia(
+      options(),
+      dependencies({
+        composePermissions: vi.fn(() => {
+          throw new PermissionRegistryError("calendar.events.read");
+        }),
+      }),
+    );
+
+    expect(check(report, "modules.permissions")).toMatchObject({
+      details: ['Permission "calendar.events.read" is declared more than once.'],
+      status: "failure",
+    });
+    expect(report.result).toBe("not-ready");
+  });
+
+  it("sanitizes an unexpected permission-composition exception to a fixed generic detail", async () => {
+    const hostile = "postgresql://admin:password@host.invalid/db /home/operator/.env";
+    const report = await diagnoseEkklesia(
+      options(),
+      dependencies({
+        composePermissions: vi.fn(() => {
+          throw new Error(hostile);
+        }),
+      }),
+    );
+
+    expect(check(report, "modules.permissions")).toMatchObject({
+      details: ["The installed permission declarations could not be composed safely."],
+      status: "failure",
+    });
+    expect(JSON.stringify(report)).not.toContain(hostile);
+  });
+
+  it("continues independent schema, migration, and PostgreSQL checks after permission composition fails", async () => {
+    const planMigrations = vi.fn(async () => plan());
+    const report = await diagnoseEkklesia(
+      options("postgresql://not-rendered.invalid/ekklesia"),
+      dependencies({
+        composePermissions: vi.fn(() => {
+          throw new PermissionRegistryError("calendar.events.read");
+        }),
+        planMigrations,
+      }),
+    );
+
+    expect(check(report, "modules.permissions").status).toBe("failure");
+    expect(check(report, "schemas.discovery").status).toBe("pass");
+    expect(check(report, "migrations.discovery").status).toBe("pass");
+    expect(check(report, "postgresql.connection").status).toBe("pass");
+    expect(planMigrations).toHaveBeenCalledOnce();
+  });
+
   it("continues independent migration and PostgreSQL checks after schema failure", async () => {
     const planMigrations = vi.fn(async () => plan());
     const report = await diagnoseEkklesia(
diff --git a/core/tests/module-discovery.test.ts b/core/tests/module-discovery.test.ts
index 2db18e9..772750e 100644
--- a/core/tests/module-discovery.test.ts
+++ b/core/tests/module-discovery.test.ts
@@ -31,7 +31,7 @@ async function createModulesDirectory(): Promise<string> {
 
 function manifestFor(id: string): Record<string, unknown> {
   return {
-    manifestVersion: 2,
+    manifestVersion: 3,
     id,
     name: `${id} module`,
     version: "0.1.0",
@@ -40,7 +40,9 @@ function manifestFor(id: string): Record<string, unknown> {
       moduleSdk: "^0.1.0",
     },
     capabilities: [{ id: `${id}.records`, version: "1.0.0" }],
-    permissions: [`${id}.records.read`],
+    permissions: [
+      { id: `${id}.records.read`, baseline: "restricted", description: "Read records." },
+    ],
     entrypoints: {
       server: "./dist/server/index.js",
     },
@@ -151,6 +153,16 @@ describe("module discovery", () => {
     expect(Object.isFrozen(modules[1]?.manifest)).toBe(true);
     const mutableBuckets = modules[1]?.manifest.storage.buckets as unknown as Array<{ id: string }>;
     expect(() => mutableBuckets.push({ id: "tampered" })).toThrow(TypeError);
+
+    expect(modules[1]?.manifest.permissions).toEqual([
+      { id: "calendar.records.read", baseline: "restricted", description: "Read records." },
+    ]);
+    expect(Object.isFrozen(modules[1]?.manifest.permissions)).toBe(true);
+    expect(Object.isFrozen(modules[1]?.manifest.permissions[0])).toBe(true);
+    const mutablePermissions = modules[1]?.manifest.permissions as unknown as Array<{
+      id: string;
+    }>;
+    expect(() => mutablePermissions.push({ id: "tampered" })).toThrow(TypeError);
   });
 
   it("aggregates malformed packages and manifests without partial discovery", async () => {
@@ -215,7 +227,9 @@ describe("module discovery", () => {
         ...manifestFor("one"),
         id: "shared",
         capabilities: [{ id: "shared.records", version: "1.0.0" }],
-        permissions: ["shared.records.read"],
+        permissions: [
+          { id: "shared.records.read", baseline: "restricted", description: "Read records." },
+        ],
       },
       packageJson: {
         name: "@test/duplicate",
@@ -229,7 +243,9 @@ describe("module discovery", () => {
         ...manifestFor("two"),
         id: "shared",
         capabilities: [{ id: "shared.records", version: "1.0.0" }],
-        permissions: ["shared.records.read"],
+        permissions: [
+          { id: "shared.records.read", baseline: "restricted", description: "Read records." },
+        ],
       },
       packageJson: {
         name: "@test/duplicate",
diff --git a/core/tests/module-migration-discovery.test.ts b/core/tests/module-migration-discovery.test.ts
index 3f22127..58d47e4 100644
--- a/core/tests/module-migration-discovery.test.ts
+++ b/core/tests/module-migration-discovery.test.ts
@@ -65,7 +65,7 @@ function manifestFor(
   } = {},
 ): Record<string, unknown> {
   return {
-    manifestVersion: 2,
+    manifestVersion: 3,
     id,
     name: `${id} module`,
     version: "0.1.0",
diff --git a/core/tests/module-permissions.test.ts b/core/tests/module-permissions.test.ts
new file mode 100644
index 0000000..8652c82
--- /dev/null
+++ b/core/tests/module-permissions.test.ts
@@ -0,0 +1,223 @@
+import { type PermissionBaseline, parseModuleManifest } from "@ekklesia/contracts";
+import { describe, expect, it } from "vitest";
+import {
+  CORE_AUTHORIZATION_GRANTS_MANAGE,
+  CORE_PERMISSION_DECLARATIONS,
+  PermissionRegistryError,
+} from "../src/authorization/index.js";
+import type { DiscoveredModule } from "../src/modules/discovery.js";
+import {
+  composePlatformPermissionRegistry,
+  ModulePermissionCompositionError,
+} from "../src/modules/permissions.js";
+
+interface PermissionInput {
+  readonly id: string;
+  readonly baseline: PermissionBaseline;
+  readonly description: string;
+}
+
+/** A minimal, in-memory `DiscoveredModule` double — no filesystem, no real discovery. The manifest
+ * is still built through `parseModuleManifest`, so a fixture that would not actually pass manifest
+ * validation cannot silently misrepresent what real discovery would ever produce. Only `manifest`
+ * is read by `composePlatformPermissionRegistry`; the rest exist only to satisfy the type. */
+function fakeModule(id: string, permissions: readonly PermissionInput[] = []): DiscoveredModule {
+  const manifest = parseModuleManifest({
+    manifestVersion: 3,
+    id,
+    name: `${id} module`,
+    version: "0.1.0",
+    compatibility: { platform: "^0.1.0", moduleSdk: "^0.1.0" },
+    permissions,
+    entrypoints: {
+      // Deliberately a path nothing will ever resolve, proving composition never imports or
+      // otherwise touches a module's server entrypoint.
+      server: "./dist/does-not-exist/index.js",
+    },
+  });
+  return {
+    directory: `/modules/${id}`,
+    entrypoints: { server: `/modules/${id}/dist/does-not-exist/index.js` },
+    manifest,
+    manifestPath: `/modules/${id}/ekklesia.module.json`,
+    packageName: `@test/${id}`,
+  };
+}
+
+describe("composePlatformPermissionRegistry", () => {
+  it("composes a Core-only registry when no modules are supplied", () => {
+    const registry = composePlatformPermissionRegistry([]);
+
+    expect(registry.declarations).toEqual(CORE_PERMISSION_DECLARATIONS);
+    expect(registry.has(CORE_AUTHORIZATION_GRANTS_MANAGE)).toBe(true);
+  });
+
+  it("composes Core and multiple discovered modules' declarations into one registry", () => {
+    const calendar = fakeModule("calendar", [
+      { id: "calendar.events.read", baseline: "restricted", description: "Read events." },
+    ]);
+    const people = fakeModule("people", [
+      { id: "people.directory.read", baseline: "authenticated", description: "Read directory." },
+    ]);
+
+    const registry = composePlatformPermissionRegistry([calendar, people]);
+
+    expect(registry.has(CORE_AUTHORIZATION_GRANTS_MANAGE)).toBe(true);
+    expect(registry.has("calendar.events.read")).toBe(true);
+    expect(registry.has("people.directory.read")).toBe(true);
+    expect(registry.declarations).toHaveLength(3);
+  });
+
+  it("preserves deterministic order: Core first, then modules in the supplied order, then declarations in manifest order", () => {
+    const calendar = fakeModule("calendar", [
+      { id: "calendar.events.read", baseline: "restricted", description: "Read events." },
+      { id: "calendar.events.write", baseline: "restricted", description: "Write events." },
+    ]);
+    const people = fakeModule("people", [
+      { id: "people.directory.read", baseline: "authenticated", description: "Read directory." },
+    ]);
+
+    const registry = composePlatformPermissionRegistry([calendar, people]);
+
+    expect(registry.declarations.map((declaration) => declaration.id)).toEqual([
+      CORE_AUTHORIZATION_GRANTS_MANAGE,
+      "calendar.events.read",
+      "calendar.events.write",
+      "people.directory.read",
+    ]);
+  });
+
+  it.each([
+    "public",
+    "authenticated",
+    "restricted",
+  ] as const)("preserves the exact %s baseline and description of a module declaration", (baseline) => {
+    const calendar = fakeModule("calendar", [
+      { id: "calendar.events.read", baseline, description: "Read calendar events." },
+    ]);
+
+    const registry = composePlatformPermissionRegistry([calendar]);
+
+    expect(registry.get("calendar.events.read")).toEqual({
+      id: "calendar.events.read",
+      baseline,
+      description: "Read calendar events.",
+    });
+  });
+
+  it("still fails closed for a permission no module or Core ever declared", () => {
+    const registry = composePlatformPermissionRegistry([]);
+
+    expect(registry.has("calendar.events.read")).toBe(false);
+    expect(registry.get("calendar.events.read")).toBeUndefined();
+  });
+
+  it("fails composition on an exact duplicate permission id supplied by two module entries", () => {
+    // A permission id must start with its own module's id, so a genuine cross-module collision is
+    // only reachable if the supplied list itself contains the same module id twice — real
+    // discovery already guarantees module-id uniqueness (`MODULE_ID_DUPLICATE`), but composition
+    // re-derives duplicate detection from the flat declaration list regardless, rather than
+    // trusting that upstream guarantee alone.
+    const first = fakeModule("calendar", [
+      { id: "calendar.events.read", baseline: "restricted", description: "Read events." },
+    ]);
+    const second = fakeModule("calendar", [
+      { id: "calendar.events.read", baseline: "public", description: "Also read events." },
+    ]);
+
+    expect(() => composePlatformPermissionRegistry([first, second])).toThrow(
+      PermissionRegistryError,
+    );
+  });
+
+  it("rejects a module contributing a permission in the reserved core.* namespace, without leaking a module path", () => {
+    // Only a module whose own id is literally "core" can produce a "core."-prefixed permission
+    // that still passes the manifest's own module-namespace check — this proves composition
+    // catches that case rather than trusting the manifest's self-consistency alone.
+    const impostor = fakeModule("core", [
+      { id: "core.rogue.permission", baseline: "restricted", description: "Not really Core's." },
+    ]);
+
+    let captured: unknown;
+    try {
+      composePlatformPermissionRegistry([impostor]);
+    } catch (error) {
+      captured = error;
+    }
+
+    expect(captured).toBeInstanceOf(ModulePermissionCompositionError);
+    expect(captured).toMatchObject({
+      code: "MODULE_PERMISSION_RESERVED_NAMESPACE",
+      moduleId: "core",
+      permissionId: "core.rogue.permission",
+    });
+    const publicText = String(captured);
+    expect(publicText).not.toContain("/modules/core");
+    expect(publicText).not.toContain("ekklesia.module.json");
+  });
+
+  it("rejects a reserved-namespace collision even when it would not otherwise duplicate an existing id", () => {
+    const impostor = fakeModule("core", [
+      {
+        id: "core.authorization.grants.manage",
+        baseline: "restricted",
+        description: "Attempting to shadow Core's own permission.",
+      },
+    ]);
+
+    expect(() => composePlatformPermissionRegistry([impostor])).toThrow(
+      ModulePermissionCompositionError,
+    );
+  });
+
+  it("removes a module's permissions from the resulting registry once that module is absent from the supplied list", () => {
+    const calendar = fakeModule("calendar", [
+      { id: "calendar.events.read", baseline: "restricted", description: "Read events." },
+    ]);
+
+    const withCalendar = composePlatformPermissionRegistry([calendar]);
+    expect(withCalendar.has("calendar.events.read")).toBe(true);
+
+    const withoutCalendar = composePlatformPermissionRegistry([]);
+    expect(withoutCalendar.has("calendar.events.read")).toBe(false);
+  });
+
+  it("performs no I/O: composition is synchronous and never imports a module's entrypoint", () => {
+    const calendar = fakeModule("calendar", [
+      { id: "calendar.events.read", baseline: "restricted", description: "Read events." },
+    ]);
+
+    // A synchronous return (not a Promise) from a fixture whose entrypoint deliberately resolves
+    // to nothing on disk is direct evidence no filesystem or module-code operation occurred.
+    const result = composePlatformPermissionRegistry([calendar]);
+    expect(result).not.toBeInstanceOf(Promise);
+    expect(result.has("calendar.events.read")).toBe(true);
+  });
+
+  it("does not mutate the supplied module list or Core's own declaration list", () => {
+    const calendar = fakeModule("calendar", [
+      { id: "calendar.events.read", baseline: "restricted", description: "Read events." },
+    ]);
+    const modules = Object.freeze([calendar]);
+    const coreDeclarationsBefore = [...CORE_PERMISSION_DECLARATIONS];
+
+    composePlatformPermissionRegistry(modules);
+
+    expect(modules).toEqual([calendar]);
+    expect(CORE_PERMISSION_DECLARATIONS).toEqual(coreDeclarationsBefore);
+  });
+
+  it("returns a registry and declarations that remain immutable, including module-contributed ones", () => {
+    const calendar = fakeModule("calendar", [
+      { id: "calendar.events.read", baseline: "restricted", description: "Read events." },
+    ]);
+
+    const registry = composePlatformPermissionRegistry([calendar]);
+
+    expect(Object.isFrozen(registry)).toBe(true);
+    expect(Object.isFrozen(registry.declarations)).toBe(true);
+    for (const declaration of registry.declarations) {
+      expect(Object.isFrozen(declaration)).toBe(true);
+    }
+  });
+});
diff --git a/core/tests/module-runtime.test.ts b/core/tests/module-runtime.test.ts
index 71edd82..28b02d9 100644
--- a/core/tests/module-runtime.test.ts
+++ b/core/tests/module-runtime.test.ts
@@ -43,7 +43,7 @@ function manifestFor(
   } = {},
 ): Record<string, unknown> {
   return {
-    manifestVersion: 2,
+    manifestVersion: 3,
     id,
     name: `${id} module`,
     version: "0.1.0",
diff --git a/core/tests/module-schema-discovery.test.ts b/core/tests/module-schema-discovery.test.ts
index e7374df..1a760f9 100644
--- a/core/tests/module-schema-discovery.test.ts
+++ b/core/tests/module-schema-discovery.test.ts
@@ -41,7 +41,7 @@ function manifestFor(
   options: { dependencies?: Record<string, unknown>; schema?: string | false } = {},
 ): Record<string, unknown> {
   return {
-    manifestVersion: 2,
+    manifestVersion: 3,
     id,
     name: `${id} module`,
     version: "0.1.0",
diff --git a/docs/architecture/decisions/0006-versioned-capabilities-and-compiled-module-runtime.md b/docs/architecture/decisions/0006-versioned-capabilities-and-compiled-module-runtime.md
index ac88c1f..0c3e073 100644
--- a/docs/architecture/decisions/0006-versioned-capabilities-and-compiled-module-runtime.md
+++ b/docs/architecture/decisions/0006-versioned-capabilities-and-compiled-module-runtime.md
@@ -19,6 +19,15 @@ registry, which conflicts with install-by-cloning discovery.
 
 ## Decision
 
+> **Partial correction — see [ADR 0014](0014-module-permission-declarations-and-platform-registry.md).**
+> `manifestVersion: 2` and the permission-string-array shape described immediately below (the
+> `permissions` field of the static manifest) were superseded by ADR 0014's `manifestVersion: 3`,
+> which represents each permission as a complete `{id, baseline, description}` declaration instead
+> of a bare identifier string. Every other decision in this ADR — versioned capabilities,
+> `defineModuleRuntime`, the compiled-runtime production loading boundary, and registration/lifecycle
+> ordering — remains accepted and unchanged. This section is left as originally accepted, as the
+> historical record of the manifest's prior shape.
+
 `ekklesia.module.json` remains the one canonical static manifest and moves to
 `manifestVersion: 2`. Provided capabilities are declarations containing an identifier and an exact
 semantic version. Required and optional capability dependencies map identifiers to semantic-version
diff --git a/docs/architecture/decisions/0013-core-authorization-and-owner-bootstrap.md b/docs/architecture/decisions/0013-core-authorization-and-owner-bootstrap.md
index 5f84daa..8634fee 100644
--- a/docs/architecture/decisions/0013-core-authorization-and-owner-bootstrap.md
+++ b/docs/architecture/decisions/0013-core-authorization-and-owner-bootstrap.md
@@ -194,6 +194,13 @@ module permission discovery/registration (a mechanism for a module to contribute
 declarations to a combined permission registry) are all deliberately out of scope for Phase 4C and
 remain later work.
 
+> **Update — see [ADR 0014](0014-module-permission-declarations-and-platform-registry.md).** Module
+> permission discovery/registration, deferred above, was implemented by ADR 0014: a module manifest
+> declares permissions in the same `PermissionDeclaration` shape this ADR's registry uses, and
+> `apps/api` composes Core's declarations with every installed module's into the one registry every
+> request is enforced against. Roles/groups/teams/tenants, owner-management HTTP, and service
+> accounts/API tokens remain deferred.
+
 ## Alternatives
 
 - Representing platform-owner status as a special grant row (a synthetic `*` permission, or a
diff --git a/docs/architecture/decisions/0014-module-permission-declarations-and-platform-registry.md b/docs/architecture/decisions/0014-module-permission-declarations-and-platform-registry.md
new file mode 100644
index 0000000..100a1ee
--- /dev/null
+++ b/docs/architecture/decisions/0014-module-permission-declarations-and-platform-registry.md
@@ -0,0 +1,233 @@
+# 0014: Module permission declarations and platform permission registry
+
+- Status: Accepted
+- Date: 2026-07-24
+- Extends: [0013](0013-core-authorization-and-owner-bootstrap.md)
+- Supersedes (partial): [0006](0006-versioned-capabilities-and-compiled-module-runtime.md) — only
+  its `manifestVersion: 2` and bare-permission-string decision. Capabilities, the compiled-runtime
+  production loading boundary, and registration/lifecycle ordering established in ADR 0006 remain
+  accepted and unchanged.
+
+## Context
+
+ADR 0013 established Core's authorization foundation — schema, evaluator, store, permission
+registry, platform-owner bootstrap, and HTTP enforcement — but explicitly deferred module
+permission registration: a module manifest (ADR 0006, `manifestVersion: 2`) could declare a
+`permissions` array, but only as bare namespaced strings. That was enough to validate a namespace
+at manifest-parse time, but not enough to *register* anything: Core's authorization registry needs
+a trusted baseline and an operator-facing description for every permission it enforces, and a bare
+string carries neither. Phase 5A closes that gap in two passes on this branch: pass 1
+(`feat: declare module permissions`) established the shared declaration contract, manifest version
+3, and a pure composition function; pass 2 (`feat: integrate module authorization`) wired the
+composed registry into `apps/api`'s production startup and its existing enforcement/explicit-grant
+HTTP surface. This ADR records the durable architecture both passes established.
+
+## Decision
+
+### `PermissionDeclaration` is the one shared, contracts-layer definition
+
+`packages/contracts/src/authorization.ts`'s `permissionDeclarationSchema` is the single
+authoritative shape of a registered permission declaration — `id` (a branded `PermissionId`),
+`baseline` (`public`/`authenticated`/`restricted`, the evaluator's trusted baseline), and
+`description` (required, trimmed, non-empty, capped at 240 characters). `.strict()` rejects any
+other property. Both Core's own authorization registry
+(`core/src/authorization/permissions.ts`) and a module manifest's `permissions` array
+(`packages/contracts/src/module-manifest.ts`) consume this exact type — Core does not declare a
+second, independently maintained interface, and a module cannot smuggle extra fields into a
+declaration Core treats as trusted. `description` is informational only: audit and operator UI
+text, never read or interpreted by the evaluator, the registry, or enforcement — changing it can
+never change what a permission allows.
+
+### Module manifests are inert JSON and the sole permission-declaration source
+
+A module's `ekklesia.module.json` remains inert, static data Core parses without executing any
+module code — consistent with ADR 0006's discovery model. Its `permissions` array is the only
+place a module declares permissions; nothing in `defineModuleRuntime` (Module SDK,
+`packages/module-sdk/src/define-module-runtime.ts`) repeats, duplicates, or otherwise re-declares
+permissions, baselines, or descriptions — exactly like it already does not repeat capabilities,
+dependencies, storage requests, or entrypoints (ADR 0006). Core never imports or runs a module's
+server entrypoint, job, or lifecycle hook to discover its permissions; only the manifest is read.
+
+### Manifest version 3 replaces version 2's bare permission strings
+
+Representing a permission as a complete declaration instead of a bare identifier string is an
+incompatible manifest shape change, so the canonical manifest moves to `manifestVersion: 3`
+(`z.literal(3)`). Version 2 — and any other version — fails validation outright; there is no
+compatibility adapter, no automatic conversion from a string array to declarations, and no dual
+acceptance of both shapes. A module author must rewrite `permissions` explicitly. `permissions`
+still defaults to an empty array when the field is omitted entirely, exactly as version 2 did.
+
+### Each module owns exactly its own namespace; `core.*` is reserved
+
+A permission id declared in a module's manifest must begin with `<module-id>.`, checked the same
+way a capability id already is (ADR 0006). Module ids never contain a dot (`moduleIdSchema` is
+plain kebab-case), so the only module whose manifest could ever declare a `core.`-prefixed
+permission and still pass this per-manifest namespace check is one whose own id is literally
+`"core"` — `core.*` itself remains reserved for Core's own permissions, enforced independently at
+composition time (below). Duplicate permission ids within one manifest fail manifest validation at
+the exact issue path `permissions[index].id`, mirroring the existing capability and storage-bucket
+duplicate checks.
+
+### Core and installed-module declarations compose into one immutable registry
+
+`composePlatformPermissionRegistry` (`core/src/modules/permissions.ts`) builds the platform's one
+permission registry from Core's own declarations (`CORE_PERMISSION_DECLARATIONS`) plus every
+already-discovered module's declarations. It reuses `createPermissionRegistry`
+(`core/src/authorization/permissions.ts`) rather than reimplementing registry construction or
+duplicate detection — there is no second, module-specific registry type. Order is deterministic:
+Core's declarations first, then each module in the exact order it appears in the caller-supplied
+list, then each module's own declarations in manifest-array order. The resulting registry, and
+every declaration it exposes, is immutable (`Object.freeze`), matching `createPermissionRegistry`'s
+existing guarantee. Composition is pure, synchronous, in-memory work over data the caller already
+discovered: no filesystem, environment, database, network, or module-code operation happens inside
+it.
+
+`core/src/authorization` continues to know nothing about modules — it does not import module
+discovery or runtime. `core/src/modules/permissions.ts` is the one-directional module-integration
+layer that depends on `core/src/authorization`, never the reverse.
+
+### The structural collision guarantee
+
+A discovered module contributing any `core.`-prefixed permission id fails composition with
+`ModulePermissionCompositionError` (`MODULE_PERMISSION_RESERVED_NAMESPACE`) — checked, and thrown,
+before that declaration is ever added to the list `createPermissionRegistry` sees. This is a
+deliberate architectural guarantee, not an incidental test limitation, and it follows from three
+facts holding simultaneously: real module discovery rejects a duplicate module id
+(`MODULE_ID_DUPLICATE`), so module ids are unique; a manifest's own namespace check requires every
+permission id to begin with `<module-id>.`; and a manifest already rejects a duplicate permission
+id within itself. Together, these mean a module can only ever produce a `core.`-prefixed permission
+if its own id is literally `"core"` — and when it does, that declaration is caught by the
+reserved-namespace check specifically, never by `createPermissionRegistry`'s generic duplicate
+detection (`PermissionRegistryError`, `PERMISSION_REGISTRY_DUPLICATE_DECLARATION`). The same three
+facts also mean a genuine *module-vs-module* exact permission-id collision cannot arise from a
+valid, really-discovered installed-module set at all — two different module ids can never produce
+the identical permission string. `createPermissionRegistry`'s own duplicate rejection remains in
+place regardless, as an unconditional, defense-in-depth guarantee of the underlying registry
+primitive; both Core's test suite and this composer's own tests exercise it directly with
+deliberately conflicting internal test data (not reachable through real discovery), specifically to
+prove the guarantee holds even though production traffic cannot currently trigger it. Both error
+families carry only a module id and/or permission id — never a module directory, manifest path,
+caught cause, or stack.
+
+### Module discovery and permission composition precede all database and HTTP construction
+
+`apps/api`'s `startApi()` (`apps/api/src/server.ts`) composes the platform registry
+(`composeStartupPermissionRegistry`, `apps/api/src/platform-permissions.ts`) as its first step,
+strictly before constructing the shared PostgreSQL pool, Better Auth, the authorization store and
+evaluator, the Hono application, or binding the HTTP server. A discovery or composition failure —
+an invalid manifest, an incompatible dependency graph, a reserved-namespace claim, or a duplicate
+declaration — therefore has nothing to close: it propagates directly out of `startApi()`, and
+startup never falls back to a Core-only registry. Modules are discovered exactly once for this
+composition; nothing re-discovers them a second time for permissions specifically.
+
+Repository and package-version information is derived from the installed repository's own layout —
+never `process.cwd()`, which a package-scoped test run does not set to the repository root. The
+repository root is resolved from the composing module's own compiled/source file location (a fixed
+number of directory levels up, identical whether running `apps/api/dist/*.js` or, under a test
+runner, `apps/api/src/*.ts`); the platform and Module SDK versions are read from their real
+`package.json` files, only when startup actually executes — never merely by importing the API's
+server, config, or app modules, which remain side-effect-free at import time.
+
+### API enforcement and explicit grants use one registry for Core and module permissions alike
+
+`createApiApp` requires the composed platform registry as an injected authorization option;
+`startApi` is the only production caller, and it always supplies the registry
+`composeStartupPermissionRegistry` produced. There is no `CORE_PERMISSION_REGISTRY` fallback
+anywhere in `apps/api` — not in `createApiApp`, not in `createGrantsRoutes`, not in the
+`requirePermission` middleware. `requirePermission` has no notion of "Core" versus "module"
+permissions; it resolves a route's declared `PermissionId` and its trusted baseline from whichever
+registry it is given, denies before principal resolution or evaluator access when the permission is
+unregistered, and never accepts a baseline, actor, or any other authorization-relevant field from
+client input. A module's `public`, `authenticated`, or `restricted` declaration is therefore
+enforced identically to Core's own `core.authorization.grants.manage`, through the identical
+evaluator and grant path, and — because registration and evaluator shape validation remain the two
+separate, independent fail-closed layers ADR 0013 established — a well-formed but unregistered
+permission (module or Core) still receives a sanitized `403` with no store or audit write, whether
+the request is a permission check or an explicit-grant creation attempt on `POST /api/v1/grants`.
+
+### The registry is a static startup snapshot
+
+There is no hot installation, removal, or dynamic mutation of the composed registry while `apps/api`
+is running. Installing, removing, or editing a module's manifest takes effect only on the next API
+restart. Removing a module from the next startup's discovered inventory makes its permission
+unknown to new authorization checks and new explicit grants — it denies exactly like any other
+unregistered permission — but never deletes, revokes, or otherwise modifies a historical grant or
+audit row already recorded for that permission. A previously granted, still-unrevoked explicit
+grant remains listed through `GET /api/v1/grants`, and revoking it explicitly remains possible,
+since revocation is keyed to the grant's own id and source kind, not to whether its permission is
+currently registered. What happens if the exact same permission id is later reinstalled or
+reactivated in a future startup — whether a preserved, still-active grant becomes effective again
+automatically — is **explicitly not decided by this ADR** and remains deferred; nothing here should
+be read as a guarantee either way.
+
+### Doctor and `modules:validate` compose the same inventory
+
+`pnpm doctor`'s `modules.permissions` check and `pnpm modules:validate` both call
+`composePlatformPermissionRegistry` over the module list their own already-completed discovery
+produced — neither rediscovers modules solely to check permissions. A structurally valid manifest
+that claims the `core.*` namespace, or that would produce a duplicate declaration, now fails before
+that module is considered installable (`modules:validate`) or reported healthy (`doctor`).
+
+### Test-only discovery overrides are not a production configuration surface
+
+`composeStartupPermissionRegistry`'s `modulesDirectoryOverride` and `discoveredModulesOverride`
+exist solely so tests can exercise discovery and composition against controlled fixtures or
+deliberately conflicting data — the latter specifically because a genuine module-vs-module
+duplicate is otherwise unreachable through real discovery, as established above. Neither is read
+from an environment variable, an HTTP header, or any other client- or deployment-controlled input;
+`main()`, the executable entry point, never sets either.
+
+## Alternatives
+
+- **Runtime, code-based permission registration** (a module registering permissions imperatively
+  from its server entrypoint or a registration hook) was rejected: it would require executing
+  module code to discover what a module can even offer, breaking the inert-manifest discovery
+  model ADR 0006 already established and that Doctor and `modules:validate` depend on for
+  offline, side-effect-free validation.
+- **A second, module-specific registry implementation** was rejected in favor of reusing
+  `createPermissionRegistry` — a second implementation would need to reimplement duplicate
+  detection, immutability, and fail-closed lookup independently, with no guarantee the two stayed
+  behaviorally identical.
+- **Falling back to `CORE_PERMISSION_REGISTRY` when discovery or composition fails** was rejected:
+  it would let the API start in a state where module-declared permissions silently do not exist,
+  which is indistinguishable from "no modules were ever installed" and would mask a real
+  configuration error as ordinary, healthy startup.
+- **Dynamic, mutable, or hot-reloaded registration** (updating the registry while the process
+  runs, without a restart) was rejected for this phase: it introduces concurrent-mutation and
+  partial-update hazards this phase does not need to solve, and a static snapshot is sufficient
+  for the current deployment model.
+- **Silently accepting version-2 manifests** (auto-converting a bare permission string into a
+  declaration with a synthesized baseline/description) was rejected: any synthesized baseline
+  would be a guess Core cannot trust, and the whole point of requiring a real declaration is that
+  the baseline and description come from the module author, not from Core inventing them.
+- **Deleting or automatically revoking a module's grants when the module disappears** was
+  rejected: it would destroy an audit trail that must remain queryable regardless of a module's
+  current installation state, and revocation is a deliberate, actor-attributed action elsewhere in
+  this system (ADR 0013) — an uninstall is not that action.
+- **Reading a permission's baseline from HTTP request input** (letting a caller assert which
+  baseline to check) was already rejected by ADR 0013 and remains rejected here: the baseline is
+  never anything but the registry's own trusted value for a request-supplied `PermissionId`.
+
+## Consequences
+
+A module gains a real, enforceable permission the moment its manifest is written correctly and the
+module is discovered — no separate registration step, no code to write beyond the manifest itself.
+Core's own permissions and every module's permissions share one enforcement mechanism end to end:
+one registry, one middleware, one explicit-grant HTTP API. The reserved `core.*` namespace and the
+structural collision guarantee mean a module can never accidentally or deliberately shadow a
+Core-owned permission without that attempt being caught, sanitized, and reported as exactly what it
+is. The cost is the one explicitly deferred question this ADR leaves open — reinstall/reactivation
+semantics for a permission whose module was previously removed — and the operational fact that
+every module inventory change requires an API restart to take effect, which is acceptable for the
+current deployment model but not the final word on it.
+
+## Deferred
+
+Module HTTP route-registration contracts (a module mounting its own endpoints); a
+`ModuleRuntimeContext` authorization service; relationship-derived grant creation (a roster or
+similar module operation using `sourceKind: "relationship"`, not the generic explicit-grant API);
+reinstall/reactivation policy for a permission whose owning module was previously removed;
+roles, groups, teams, and tenants; an owner-management HTTP endpoint; service accounts and API
+tokens for non-browser automation; module installation/uninstallation state and hot reload; Surfaces
+and UI; and workflow/event automation. None of these are implemented by, or required for, the
+architecture this ADR records.
diff --git a/docs/architecture/decisions/README.md b/docs/architecture/decisions/README.md
index 49a90cd..96c8c0b 100644
--- a/docs/architecture/decisions/README.md
+++ b/docs/architecture/decisions/README.md
@@ -22,3 +22,4 @@ ADRs capture durable choices that affect more than one package or phase.
 - [0011: Identity HTTP transport and public enrollment policy](0011-identity-http-transport-and-public-enrollment.md) — Accepted
 - [0012: Production-equivalent identity origin/CSRF protection under test](0012-identity-origin-csrf-protection-under-test.md) — Accepted
 - [0013: Core authorization, permission registry, and platform-owner bootstrap](0013-core-authorization-and-owner-bootstrap.md) — Accepted
+- [0014: Module permission declarations and platform permission registry](0014-module-permission-declarations-and-platform-registry.md) — Accepted
diff --git a/docs/development/api.md b/docs/development/api.md
index 699d324..36b4237 100644
--- a/docs/development/api.md
+++ b/docs/development/api.md
@@ -206,6 +206,18 @@ trusted-proxy IP policy is configured, Better Auth may fall back to a single sha
 rather than a genuine per-client bucket. This is not yet the final, distributed, production-grade
 abuse-control design.
 
+## Module permission discovery at startup
+
+Before any database pool, Better Auth instance, or HTTP server exists, `startApi()` discovers
+every installed module's manifest and composes the one immutable platform permission registry
+(Core's own declarations plus every module's) that the rest of startup — and every subsequent
+request — enforces against. A discovery or composition failure (an invalid manifest, a module
+claiming the reserved `core.*` permission namespace, a duplicate permission declaration, ...) fails
+startup outright, before there is anything to close; startup never falls back to a Core-only
+registry. See [Authorization](authorization.md#api-startup-discovery-composition-and-enforcement)
+for the full flow, and [Module development](module-development/README.md) for the manifest shape
+that produces these declarations.
+
 ## Database connection limits
 
 `apps/api/src/server.ts` configures the one shared PostgreSQL pool (`createApiDatabase`,
diff --git a/docs/development/authorization.md b/docs/development/authorization.md
index 177eda3..7bc9243 100644
--- a/docs/development/authorization.md
+++ b/docs/development/authorization.md
@@ -1,24 +1,29 @@
 # Authorization
 
-Phase 4C establishes Core's authorization foundation and, in a second slice, wires it into
-`apps/api` over HTTP. This page covers both: the schema/evaluator/store/registry Core owns, the
-platform-owner bootstrap operation, and the enforcement/endpoints `apps/api` exposes on top of it.
-See [Identity](identity.md) for the identity foundation this builds on, [API](api.md) for the
-HTTP transport it is mounted into, and [ADR
-0013](../architecture/decisions/0013-core-authorization-and-owner-bootstrap.md) for the durable
-architectural decisions this page describes in more implementation detail.
+Phase 4C establishes Core's authorization foundation and wires it into `apps/api` over HTTP. Phase
+5A extends that same system so an installed module's declared permissions are enforced and
+grantable through the identical mechanism as Core's own. This page covers all of it: the
+schema/evaluator/store/registry Core owns, the platform-owner bootstrap operation, the
+enforcement/endpoints `apps/api` exposes, and how module manifests contribute to the one platform
+permission registry. See [Identity](identity.md) for the identity foundation this builds on,
+[API](api.md) for the HTTP transport it is mounted into, [Module development](../module-development/README.md)
+for the manifest shape that declares module permissions, and
+[ADR 0013](../architecture/decisions/0013-core-authorization-and-owner-bootstrap.md) /
+[ADR 0014](../architecture/decisions/0014-module-permission-declarations-and-platform-registry.md)
+for the durable architectural decisions this page describes in more implementation detail.
 
 ## What this is, and is not
 
-After this slice, an operator can bootstrap exactly one platform owner for a deployment, a
-platform owner (or anyone they delegate to) can create, list, and revoke *explicit* authorization
-grants over HTTP, and every enforced route fails closed on infrastructure trouble instead of
-silently allowing or denying. It is **not** possible to manage roles, groups, teams, or tenants;
-there is no owner-management HTTP endpoint (revoking or re-granting platform-owner status is not
-exposed at all in this slice); there are no service accounts or API tokens, so only
-cookie-authenticated browser callers can reach the mutation routes; and relationship-derived
-grants (for example, automatic access from a roster assignment) cannot yet be created by anything
-in this repository — the generic grants API structurally cannot create or revoke them either.
+An operator can bootstrap exactly one platform owner for a deployment; a platform owner (or anyone
+they delegate to) can create, list, and revoke *explicit* authorization grants over HTTP for any
+registered permission, Core-owned or module-owned; and every enforced route fails closed on
+infrastructure trouble instead of silently allowing or denying. It is **not** possible to manage
+roles, groups, teams, or tenants; there is no owner-management HTTP endpoint (revoking or
+re-granting platform-owner status is not exposed at all); there are no service accounts or API
+tokens, so only cookie-authenticated browser callers can reach the mutation routes; a module cannot
+yet mount its own HTTP routes; and relationship-derived grants (for example, automatic access from
+a roster assignment) cannot yet be created by anything in this repository — the generic grants API
+structurally cannot create or revoke them either.
 
 ## Core authorization schema and store
 
@@ -31,10 +36,19 @@ permission registry, the bootstrap operation, and HTTP enforcement.
 
 ## Permission registry
 
-`core/src/authorization/permissions.ts` defines the smallest possible registry: an immutable,
-composition-time-built `PermissionRegistry` distinguishing a *registered declaration*
-(`{id, baseline, description}`) from a merely well-formed `PermissionId` string. Two properties
-matter:
+`PermissionDeclaration` (`packages/contracts/src/authorization.ts`) is the one authoritative,
+strict Zod-validated shape of a *registered declaration* — exactly `{id, baseline, description}`,
+nothing more — shared by Core's authorization registry and a module manifest's `permissions` array.
+Core does not declare its own separate interface for this; `core/src/authorization/permissions.ts`
+imports and re-exports the contracts type so existing `@ekklesia/core/authorization` consumers keep
+importing it from the same place. `id` is a branded `PermissionId`, `baseline` is
+`public`/`authenticated`/`restricted`, and `description` is required, trimmed, non-empty operator
+and audit-facing text up to 240 characters — informational only; nothing in authorization reads or
+interprets it.
+
+`core/src/authorization/permissions.ts` defines the smallest possible registry on top of that
+shape: an immutable, composition-time-built `PermissionRegistry` distinguishing a *registered*
+declaration from a merely well-formed `PermissionId` string. Two properties matter:
 
 - **Duplicate declarations fail at composition time.** `createPermissionRegistry` throws
   `PermissionRegistryError` (`PERMISSION_REGISTRY_DUPLICATE_DECLARATION`) if the same permission id
@@ -44,14 +58,121 @@ matter:
   and return `undefined`/`false` for anything malformed or simply not registered — there is no
   "trust this permission id because it looks right" path.
 
-`CORE_PERMISSION_REGISTRY` is Core's own registry, currently declaring exactly one permission:
+Permission matching is always exact string equality, both in the registry and in the evaluator's
+grant lookup — there is no wildcard, prefix, or namespace-inheritance matching anywhere; the
+required dot-namespacing in a `PermissionId` (`core.authorization.grants.manage`,
+`calendar.events.read`) is purely a naming convention for readability and collision-avoidance.
+
+`CORE_PERMISSION_REGISTRY` is Core's own registry, currently declaring exactly one permission, and
+remains what Core-only code and existing tests use directly. `CORE_PERMISSION_DECLARATIONS` is the
+same declarations exported separately in deeply immutable form, for the module composition helper
+below to build on without reconstructing Core's own list:
 
 | Permission | Baseline | Description |
 | --- | --- | --- |
 | `core.authorization.grants.manage` | `restricted` | Create, list, and revoke another user's explicit authorization grants. |
 
-Module-owned permission declarations and a discovery/composition path for them are out of scope
-for this slice (deferred to the Module SDK work).
+## Module permission composition
+
+A module manifest's `permissions` array (`manifestVersion: 3`,
+[Module development](../module-development/README.md#canonical-manifest)) declares permissions in
+the exact same `PermissionDeclaration` shape — but declaring one there only proves it is
+well-formed and correctly namespaced under the module's own id; it does not, by itself, register
+anything. `composePlatformPermissionRegistry` (`core/src/modules/permissions.ts`) is the trusted
+composer: given an already-discovered module list (for example `discoverModules`'s result), it
+builds one combined `PermissionRegistry` from `CORE_PERMISSION_DECLARATIONS` followed by every
+module's declarations, in discovery order, using the same `createPermissionRegistry` constructor —
+not a second registry implementation. Declaration order is deterministic: Core first, then each
+module in the order it appears in the supplied list, then each module's own declarations in
+manifest-array order. A duplicate exact permission id anywhere in that combined list fails
+composition the same way a duplicate would fail `createPermissionRegistry` directly — since a
+permission id must start with its own module's id, this is only reachable across two entries that
+share the same module id, which real discovery already rules out (`MODULE_ID_DUPLICATE`);
+composition still re-derives the check rather than trusting that guarantee alone.
+
+`core.*` is reserved for Core's own permissions. A discovered module contributing any
+`core.`-prefixed permission id fails composition with `ModulePermissionCompositionError`
+(`MODULE_PERMISSION_RESERVED_NAMESPACE`) before duplicate detection ever runs — since every
+Core-owned id is itself `core.`-prefixed, this is what actually catches a module attempting to
+collide with a Core-owned id specifically, regardless of that module's own id and regardless of
+whether the id happens to exactly match an existing declaration.
+
+Composition itself performs no filesystem, environment, database, network, or module-code
+operation — it is pure, synchronous, in-memory composition over data a caller already discovered
+and validated (`core/src/authorization` and `core/src/modules` remain independent of each other in
+the other direction: authorization never imports module discovery or runtime, only the module
+integration layer depends on authorization).
+
+## API startup: discovery, composition, and enforcement
+
+`apps/api` builds its production permission registry from the complete installed-module inventory
+every time it starts, in a fixed order, before anything else:
+
+1. Resolve module-discovery configuration (the repository's `modules/` directory, and the
+   platform/Module SDK versions from their own `package.json` files).
+2. Discover installed modules (`discoverModules`).
+3. Compose the platform permission registry (`composePlatformPermissionRegistry`).
+4. Construct the shared PostgreSQL pool.
+5. Construct Identity and Authorization.
+6. Create the Hono application, injecting the composed registry.
+7. Bind the HTTP server.
+
+`apps/api/src/platform-permissions.ts`'s `composeStartupPermissionRegistry` performs steps 1–3.
+Discovery and composition happen strictly before any pool, Better Auth instance, or HTTP server
+exists, so a failure there — an invalid manifest, a broken dependency graph, a module claiming the
+reserved `core.*` namespace, or a duplicate permission declaration — leaves nothing to close: it
+propagates directly out of `startApi()`, and startup never falls back to `CORE_PERMISSION_REGISTRY`
+or any Core-only registry. The repository root is resolved from this compiled/source file's own
+location (two directories up from `apps/api/dist/*.js` or `apps/api/src/*.ts`, which sit at
+identical depth), never `process.cwd()` — a package-scoped test run has `apps/api` as its working
+directory, which would otherwise point discovery at the wrong `modules/` directory.
+
+The resulting registry is a **static, immutable snapshot** taken once at startup. There is no hot
+installation, removal, or dynamic mutation of installed modules or their permissions while the
+process is running — an installed-module change (add, remove, or edit a manifest) takes effect
+only on the next API restart. `createApiApp` requires this registry as part of its injected
+authorization options; `requirePermission` and the explicit-grant routes
+(`GET/POST /api/v1/grants`, `POST /api/v1/grants/:grantId/revoke`) consult exactly the same
+registry Core's own `core.authorization.grants.manage` permission is registered in, so a
+module-owned permission and Core's own are enforced, and grantable, through the identical
+mechanism — there is no module-specific enforcement path. Removing a module from the next
+startup's discovered inventory makes its permission unknown to new authorization checks and new
+grants (a well-formed but unregistered permission denies exactly as before, including on
+`POST /api/v1/grants`, which checks the registry explicitly before calling the store); it never
+deletes, revokes, or otherwise modifies a historical grant or audit row already recorded for that
+permission. A previously granted, still-unrevoked explicit grant for a now-uninstalled module's
+permission remains visible through `GET /api/v1/grants` and simply stops being something a
+`restricted`-baseline authorization check can newly succeed against. Listing and explicit
+revocation (`POST /api/v1/grants/:grantId/revoke`) are unaffected either way: neither route
+consults the registry for a grant that already exists — revocation is keyed to the grant's own id
+and source kind, not to whether its permission is currently registered — so a manager can still
+revoke a historical grant for a now-unregistered permission.
+
+Module HTTP route registration (a module owning its own mounted endpoints) and relationship-derived
+grant creation both remain deferred, unrelated slices — this work only makes a module's *declared*
+permission usable through Core's existing, generic explicit-grant HTTP API and enforcement
+middleware.
+
+## Doctor and modules:validate
+
+`pnpm doctor`'s `modules.permissions` check composes the platform permission registry from the
+same already-discovered module list `modules.discovery`/`modules.dependencies` already produced —
+it never re-discovers modules itself. A successful composition reports a pass with a safe
+declaration count (for example `1 permission registered` with zero installed modules, since Core's
+own declaration is always present); zero installed modules still composes successfully. If module
+discovery itself failed, the check is skipped, matching the existing schema/migration checks'
+pattern; if discovery succeeded but composition failed, schema and migration checks still run
+normally — they are independent diagnostics that do not depend on the permission registry. A
+reserved-namespace violation (`ModulePermissionCompositionError`,
+`MODULE_PERMISSION_RESERVED_NAMESPACE`) and a generic duplicate declaration
+(`PermissionRegistryError`, `PERMISSION_REGISTRY_DUPLICATE_DECLARATION`) are reported as distinct
+failure details — a reserved `core.*` claim is never described as an ordinary duplicate, even
+though (see above) a module attempting to exactly collide with a Core-owned id always surfaces as
+the reserved-namespace failure specifically. Composition requires no database and never imports or
+executes a module's server entrypoint. `pnpm modules:validate` composes the same registry from its
+own already-discovered modules for the identical reason: a manifest that is otherwise structurally
+valid but claims the `core.*` namespace, or duplicates a permission id, now fails validation before
+that module is considered installable.
 
 ### Two layers, not one: shape versus registration
 
@@ -218,7 +339,7 @@ cannot smuggle a server-derived field through a request that otherwise looks wel
 | Field | Required | Notes |
 | --- | --- | --- |
 | `subjectUserId` | Yes | Must parse as a valid `EkklesiaUserId`. |
-| `permission` | Yes | Must parse as a valid `PermissionId` **and** be registered in `CORE_PERMISSION_REGISTRY` — see [Two layers, not one](#two-layers-not-one-shape-versus-registration). A malformed value is a `400`; a well-formed but unregistered one is a `403`. |
+| `permission` | Yes | Must parse as a valid `PermissionId` **and** be registered in the running API's composed platform registry (Core's own declarations plus every installed module's — see [API startup](#api-startup-discovery-composition-and-enforcement)). A malformed value is a `400`; a well-formed but unregistered one is a `403`. |
 | `scope` | No | Must parse as a valid, non-empty `AuthorizationScope` if present. |
 | `expiresAt` | No | An ISO date string if present. |
 
@@ -287,7 +408,19 @@ there are or whether the failure happened during startup.
 ## Tests
 
 - `core/tests/authorization-permissions.test.ts` — registry construction, duplicate-declaration
-  rejection, fail-closed lookup, `CORE_PERMISSION_REGISTRY` contents.
+  rejection, fail-closed lookup, `CORE_PERMISSION_REGISTRY`/`CORE_PERMISSION_DECLARATIONS` contents
+  and immutability.
+- `packages/contracts/tests/authorization.test.ts` — `PermissionDeclaration` shape validation:
+  every baseline, unknown-property/missing-baseline/invalid-baseline/blank-or-oversized-description
+  rejection, malformed permission id rejection.
+- `packages/contracts/tests/module-manifest.test.ts` — `manifestVersion: 3` acceptance and version
+  2/string-array rejection, per-declaration validation (mirroring the contracts-level cases above),
+  module-namespace enforcement, duplicate permission id at its exact `permissions[index].id` issue
+  path, and default-empty-list isolation across separate parses.
+- `core/tests/module-permissions.test.ts` — `composePlatformPermissionRegistry`: Core-only and
+  Core-plus-modules composition, deterministic ordering, exact baseline/description preservation,
+  duplicate rejection, reserved `core.*` namespace rejection, unknown-permission denial once a
+  module is absent from the supplied list, and input/output immutability with no I/O.
 - `core/tests/database-pool.test.ts` — shared pool validation/construction (mirrors the existing
   per-domain database tests).
 - `core/tests/authorization-bootstrap.test.ts` — import safety, `PlatformOwnerBootstrapError`
@@ -298,13 +431,27 @@ there are or whether the failure happened during startup.
   its audit row, nonexistent-user rollback, second-bootstrap rejection, a revoked historical owner
   still blocking reuse, and the concurrency scenario described above.
 - `apps/api/tests/authorization-middleware.test.ts` — allow/deny/unregistered/infrastructure-
-  failure paths for `requirePermission`, and `requireTrustedOrigin`'s allow/reject behavior.
+  failure paths for `requirePermission`, and `requireTrustedOrigin`'s allow/reject behavior;
+  a dedicated block proves `requirePermission` enforces module-owned `public`/`authenticated`/
+  `restricted` declarations identically to Core's own, that the evaluator receives the registry's
+  exact id/baseline, and that an absent module permission denies before principal resolution or
+  evaluator access — the shared mechanism, not a module-specific code path.
+- `core/tests/doctor.test.ts` / `core/tests/doctor-report.test.ts` — the `modules.permissions`
+  check: passing count (including zero installed modules), reuse of the already-discovered module
+  list (no rediscovery), distinct reserved-namespace vs. generic-duplicate failure detail, an
+  unexpected-exception fallback to a fixed generic detail, schema/migration checks continuing
+  independently after a composition failure, and the exact documented terminal report shape.
 - `apps/api/tests/grants-routes.test.ts` — all three routes' allow/deny paths, disallowed-field
   and malformed-field rejection, a well-formed but unregistered target permission rejected with
   `403` and `store.createGrant` never called, exact server-derived-argument assertions,
   sourceKind-filtered revoke returning 404 for a non-explicit grant, origin enforcement, and
   store-error mapping.
-- `apps/api/tests/lifecycle.test.ts` — shared-pool-closes-exactly-once coverage described above.
+- `apps/api/tests/lifecycle.test.ts` — shared-pool-closes-exactly-once coverage described above,
+  plus module-permission composition at startup: discovery/composition run before any pool,
+  Better Auth instance, or HTTP server is constructed; a discovery failure, a reserved-namespace
+  failure, and a generic duplicate-declaration failure each create zero pools/Better Auth
+  instances/servers and remain distinguishable from one another; zero installed modules still
+  succeeds; and import safety for `platform-permissions.ts`.
 - `core/tests/authorization-evaluator.test.ts` — includes a malformed permission id denying under
   every baseline (`public`, `authenticated`, `restricted`), never just `restricted`, and never
   consulting the grant or platform-owner ports for `public`/`authenticated`.
@@ -314,11 +461,21 @@ there are or whether the failure happened during startup.
   account then managing grants as a non-owner, a well-formed but unregistered target permission
   rejected with `403` (with no grant or audit row written) even for a caller with full
   grant-management authority, a create/list/revoke cycle with audit-trail verification (using
-  `core.authorization.grants.manage` itself as the granted permission, since module permission
-  registration does not exist yet), disallowed-field rejection, missing/untrusted-origin rejection,
-  a real relationship-derived grant proven invisible to LIST and unreachable by REVOKE, and an
-  infrastructure-failure scenario (database dropped out from under a live session) asserting a
-  `503` rather than a false `403` or `200`.
+  `core.authorization.grants.manage` itself as the granted permission), disallowed-field
+  rejection, missing/untrusted-origin rejection, a real relationship-derived grant proven
+  invisible to LIST and unreachable by REVOKE, and an infrastructure-failure scenario (database
+  dropped out from under a live session) asserting a `503` rather than a false `403` or `200`.
+- `apps/api/tests/api-http.integration.ts`, "installed module permission end-to-end" (live
+  PostgreSQL) — a real temporary module fixture (structurally valid manifest, a server entrypoint
+  that throws if ever imported) proves the complete chain against a real API process and real
+  Doctor run: Doctor recognizes the installed module and passes permission composition without
+  executing its entrypoint; API startup discovers the fixture and makes its declared permission
+  enforceable; a bootstrapped platform owner creates a real explicit grant for the module's
+  permission with an atomic audit row recorded under its exact declared id; a well-formed
+  permission absent from the composed registry is rejected with no grant or audit write; and,
+  after a clean shutdown and restart with the module absent, the same permission can no longer be
+  newly granted while the earlier, still-unrevoked grant remains listed — with ordinary identity
+  HTTP behavior (sign-up, sign-in, principal resolution) unaffected throughout.
 
 Run the focused suites:
 
@@ -330,7 +487,10 @@ pnpm test:migrations:integration   # requires Docker
 
 ## Next phase
 
-Phase 4C, pass 3: deferred from this slice — module discovery/Module SDK permission
-registration, roles/groups/teams/tenants, an owner-management HTTP endpoint, service
-accounts/API tokens for non-browser automation, and relationship-derived grant creation itself
-(a roster or similar module operation, not the generic grants API).
+Deliberately deferred, unrelated slices — see
+[ADR 0014](../architecture/decisions/0014-module-permission-declarations-and-platform-registry.md#deferred)
+for the complete list: module HTTP route-registration contracts (a module mounting its own
+endpoints), a module runtime-context authorization service, relationship-derived grant creation
+itself (a roster or similar module operation, not the generic grants API), reinstall/reactivation
+policy for a permission whose owning module was previously removed, roles/groups/teams/tenants, an
+owner-management HTTP endpoint, and service accounts/API tokens for non-browser automation.
diff --git a/docs/development/repository-workflow.md b/docs/development/repository-workflow.md
index 1d1e266..b1a56a8 100644
--- a/docs/development/repository-workflow.md
+++ b/docs/development/repository-workflow.md
@@ -101,6 +101,8 @@ Modules
     0 modules installed
   ✓ Dependency graph valid
     Dependency order is empty.
+  ✓ Permission registry composed
+    1 permission registered
 
 Data definitions
   ✓ Schemas validated
diff --git a/docs/module-development/README.md b/docs/module-development/README.md
index 6dcbddf..b11449f 100644
--- a/docs/module-development/README.md
+++ b/docs/module-development/README.md
@@ -48,7 +48,7 @@ without executing module code:
 
 ```json
 {
-  "manifestVersion": 2,
+  "manifestVersion": 3,
   "id": "calendar",
   "name": "Calendar",
   "version": "0.1.0",
@@ -77,7 +77,13 @@ without executing module code:
       "version": "1.0.0"
     }
   ],
-  "permissions": ["calendar.events.read"],
+  "permissions": [
+    {
+      "id": "calendar.events.read",
+      "baseline": "restricted",
+      "description": "Read calendar events."
+    }
+  ],
   "storage": {
     "buckets": [
       {
@@ -99,6 +105,33 @@ has its own exact semantic version because its public contract can evolve indepe
 module release. Capability dependencies use semantic-version ranges. Optional means the provider
 may be absent; if it is installed, its capability version must be compatible.
 
+`manifestVersion: 3` (`packages/contracts/src/module-manifest.ts`) is a required, exact literal — a
+version 2 manifest, and any other version, fails validation outright rather than being silently
+reinterpreted; there is no compatibility adapter or automatic migration between manifest versions.
+Each `permissions` entry is a complete declaration, not a bare identifier string: `id` (must begin
+with `<module-id>.`, exactly like a capability id), `baseline` (`public`, `authenticated`, or
+`restricted` — see [Authorization](../development/authorization.md#permission-registry) for what
+each baseline means to the evaluator), and a required, non-blank `description` up to 240 characters
+for operators and audit review only — Core's authorization layer never reads or interprets it.
+Declaring a permission here is necessary but not sufficient for it to be enforceable: the manifest
+only proves the declaration is well-formed and namespaced correctly, and does not itself register
+anything with Core's authorization registry (see
+[Authorization](../development/authorization.md#module-permission-composition) for how Core composes
+every discovered module's declarations, alongside its own, into the one platform permission
+registry). A duplicate permission id within one manifest, or a permission id outside the module's
+own namespace, fails validation the same way a duplicate or misnamed capability id does.
+
+`apps/api` composes this registry once, at startup, from every module installed at that time (see
+[Authorization](../development/authorization.md#api-startup-discovery-composition-and-enforcement));
+installing, removing, or editing a module's manifest takes effect only on the next API restart, not
+while it is running. Once installed and discovered, a module's declared permission is enforceable
+and grantable through exactly the same mechanism as any Core-owned permission — the same
+`requirePermission` middleware and the same `/api/v1/grants` explicit-grant HTTP API a platform
+owner (or anyone they delegate `core.authorization.grants.manage` to) already uses. Mounting a
+module's own HTTP routes, and creating relationship-derived grants automatically from a module's
+own business logic, are both still later work; this only makes a declared permission *usable*
+through Core's existing, generic surface.
+
 Server and optional schema entrypoints must be compiled ESM `.js` or `.mjs` files. The optional
 migration entrypoint is a directory. All entrypoints begin with `./`, remain inside the package,
 and exist. Production never imports source TypeScript directly. Modules with no tables omit
diff --git a/packages/contracts/src/authorization.ts b/packages/contracts/src/authorization.ts
index 85aa1e3..a1d8e03 100644
--- a/packages/contracts/src/authorization.ts
+++ b/packages/contracts/src/authorization.ts
@@ -1,4 +1,5 @@
 import { z } from "zod";
+import { permissionIdSchema } from "./identifiers.js";
 import type { EkklesiaUserId } from "./identity.js";
 import { parseEkklesiaUserId } from "./identity.js";
 
@@ -51,6 +52,44 @@ export const permissionBaselineSchema = z.enum(["public", "authenticated", "rest
 
 export type PermissionBaseline = z.infer<typeof permissionBaselineSchema>;
 
+/**
+ * The one authoritative shape of a *registered* permission declaration — what both Core's
+ * authorization registry (`core/src/authorization/permissions.ts`) and a module manifest's
+ * `permissions` array parse into, so a permission a module declares and a permission Core's
+ * registry enforces are provably the same structure, never two independently maintained
+ * interfaces that could quietly drift apart. `.strict()` rejects any property beyond these three —
+ * there is no room for a module or a future caller to smuggle extra, unvalidated metadata into a
+ * declaration Core treats as trusted.
+ *
+ * This schema says nothing about *registration*: parsing a value into a `PermissionDeclaration`
+ * only proves it is shaped correctly, not that any registry has accepted it. Registration remains
+ * exclusively `PermissionRegistry`'s responsibility (`core/src/authorization/permissions.ts`) —
+ * built by composing exactly the declarations a trusted caller supplies, Core's own plus every
+ * discovered module's.
+ */
+export const permissionDeclarationSchema = z
+  .object({
+    id: permissionIdSchema,
+    baseline: permissionBaselineSchema,
+    /** Operator/audit-facing text only. Authorization never reads, parses, or otherwise
+     * interprets this value — changing it can never change what a permission allows. */
+    description: z.string().trim().min(1).max(240),
+  })
+  .strict();
+
+export type PermissionDeclarationInput = z.input<typeof permissionDeclarationSchema>;
+export type PermissionDeclaration = z.output<typeof permissionDeclarationSchema>;
+
+export function parsePermissionDeclaration(input: unknown): PermissionDeclaration {
+  return permissionDeclarationSchema.parse(input);
+}
+
+export function safeParsePermissionDeclaration(
+  input: unknown,
+): z.ZodSafeParseResult<PermissionDeclaration> {
+  return permissionDeclarationSchema.safeParse(input);
+}
+
 /**
  * Distinguishes a grant an operator/system created directly (`explicit`) from one a module
  * created as a side effect of its own business operation, attributed to some source the module
diff --git a/packages/contracts/src/module-manifest.ts b/packages/contracts/src/module-manifest.ts
index 322939b..24310ba 100644
--- a/packages/contracts/src/module-manifest.ts
+++ b/packages/contracts/src/module-manifest.ts
@@ -1,10 +1,10 @@
 import { z } from "zod";
+import { type PermissionDeclaration, permissionDeclarationSchema } from "./authorization.js";
 import {
   capabilityIdSchema,
   logicalBucketIdSchema,
   moduleIdSchema,
   packageRelativePathSchema,
-  permissionIdSchema,
 } from "./identifiers.js";
 import { semanticVersionRangeSchema, semanticVersionSchema } from "./versions.js";
 
@@ -89,7 +89,7 @@ function reportDuplicates(
 
 export const moduleManifestSchema = z
   .object({
-    manifestVersion: z.literal(2),
+    manifestVersion: z.literal(3),
     id: moduleIdSchema,
     name: z.string().trim().min(1).max(120),
     description: z.string().trim().min(1).max(500).optional(),
@@ -102,7 +102,7 @@ export const moduleManifestSchema = z
       .strict(),
     dependencies: dependenciesSchema,
     capabilities: z.array(moduleCapabilityDeclarationSchema).default(() => []),
-    permissions: z.array(permissionIdSchema).default(() => []),
+    permissions: z.array(permissionDeclarationSchema).default(() => []),
     storage: moduleStorageSchema,
     entrypoints: entrypointsSchema,
   })
@@ -114,7 +114,12 @@ export const moduleManifestSchema = z
       context,
       "id",
     );
-    reportDuplicates(manifest.permissions, ["permissions"], context);
+    reportDuplicates(
+      manifest.permissions.map((permission) => permission.id),
+      ["permissions"],
+      context,
+      "id",
+    );
     reportDuplicates(
       manifest.storage.buckets.map((bucket) => bucket.id),
       ["storage", "buckets"],
@@ -133,11 +138,11 @@ export const moduleManifestSchema = z
     });
 
     manifest.permissions.forEach((permission, index) => {
-      if (!permission.startsWith(`${manifest.id}.`)) {
+      if (!permission.id.startsWith(`${manifest.id}.`)) {
         context.addIssue({
           code: "custom",
           message: `Permission identifiers owned by this module must begin with ${manifest.id}.`,
-          path: ["permissions", index],
+          path: ["permissions", index, "id"],
         });
       }
     });
@@ -197,3 +202,7 @@ export type ModuleManifestInput = z.input<typeof moduleManifestSchema>;
 export type ModuleManifest = z.output<typeof moduleManifestSchema>;
 export type ModuleBucketDeclaration = z.output<typeof moduleBucketDeclarationSchema>;
 export type ModuleCapabilityDeclaration = z.output<typeof moduleCapabilityDeclarationSchema>;
+/** A module manifest's `permissions` array entry — the same `PermissionDeclaration` shape Core's
+ * authorization registry consumes (`packages/contracts/src/authorization.ts`), aliased here for
+ * discoverability alongside the manifest's other per-field declaration types. */
+export type ModulePermissionDeclaration = PermissionDeclaration;
diff --git a/packages/contracts/tests/authorization.test.ts b/packages/contracts/tests/authorization.test.ts
index cd605e8..65a5a16 100644
--- a/packages/contracts/tests/authorization.test.ts
+++ b/packages/contracts/tests/authorization.test.ts
@@ -10,10 +10,13 @@ import {
   isUserActor,
   parseAuthorizationScope,
   parseGrantId,
+  parsePermissionDeclaration,
   parsePermissionId,
+  type PermissionDeclaration,
   type PermissionId,
   safeParseAuthorizationScope,
   safeParseGrantId,
+  safeParsePermissionDeclaration,
   safeParsePermissionId,
   SYSTEM_ACTOR,
 } from "../src/index.js";
@@ -50,6 +53,87 @@ describe("PermissionId", () => {
   });
 });
 
+describe("PermissionDeclaration", () => {
+  function validDeclaration(): Record<string, unknown> {
+    return {
+      id: "calendar.events.read",
+      baseline: "restricted",
+      description: "Read calendar events.",
+    };
+  }
+
+  it.each([
+    "public",
+    "authenticated",
+    "restricted",
+  ] as const)("parses a well-formed declaration with baseline %s", (baseline) => {
+    const declaration = parsePermissionDeclaration({ ...validDeclaration(), baseline });
+    expect(declaration).toEqual({
+      id: "calendar.events.read",
+      baseline,
+      description: "Read calendar events.",
+    });
+  });
+
+  it("is exactly three fields — this is the one shape both a module manifest and Core's registry parse into", () => {
+    const declaration: PermissionDeclaration = parsePermissionDeclaration(validDeclaration());
+    expect(Object.keys(declaration).sort()).toEqual(["baseline", "description", "id"]);
+  });
+
+  it("rejects an unknown declaration property", () => {
+    const result = safeParsePermissionDeclaration({
+      ...validDeclaration(),
+      scope: "calendar-event:123",
+    });
+    expect(result.success).toBe(false);
+  });
+
+  it("rejects a missing baseline", () => {
+    const { baseline: _baseline, ...withoutBaseline } = validDeclaration();
+    expect(safeParsePermissionDeclaration(withoutBaseline).success).toBe(false);
+  });
+
+  it("rejects an invalid baseline", () => {
+    expect(
+      safeParsePermissionDeclaration({ ...validDeclaration(), baseline: "superuser" }).success,
+    ).toBe(false);
+  });
+
+  it("rejects a missing, empty, or whitespace-only description", () => {
+    const { description: _description, ...withoutDescription } = validDeclaration();
+    expect(safeParsePermissionDeclaration(withoutDescription).success).toBe(false);
+    expect(safeParsePermissionDeclaration({ ...validDeclaration(), description: "" }).success).toBe(
+      false,
+    );
+    expect(
+      safeParsePermissionDeclaration({ ...validDeclaration(), description: "   " }).success,
+    ).toBe(false);
+  });
+
+  it("rejects a description beyond the fixed maximum length", () => {
+    expect(
+      safeParsePermissionDeclaration({ ...validDeclaration(), description: "x".repeat(240) })
+        .success,
+    ).toBe(true);
+    expect(
+      safeParsePermissionDeclaration({ ...validDeclaration(), description: "x".repeat(241) })
+        .success,
+    ).toBe(false);
+  });
+
+  it("rejects a malformed permission id", () => {
+    expect(
+      safeParsePermissionDeclaration({ ...validDeclaration(), id: "not-namespaced" }).success,
+    ).toBe(false);
+  });
+
+  it("never registers or evaluates anything — parsing only proves the value is well-formed", () => {
+    // Nothing here consults a registry: `parsePermissionDeclaration` succeeding says nothing
+    // about whether `calendar.events.read` is actually registered anywhere.
+    expect(() => parsePermissionDeclaration(validDeclaration())).not.toThrow();
+  });
+});
+
 describe("GrantId", () => {
   it("parses a non-empty opaque grant id and rejects empty/oversized input", () => {
     const id: GrantId = parseGrantId("11111111-1111-1111-1111-111111111111");
diff --git a/packages/contracts/tests/module-manifest.test.ts b/packages/contracts/tests/module-manifest.test.ts
index 5997339..07b5fb9 100644
--- a/packages/contracts/tests/module-manifest.test.ts
+++ b/packages/contracts/tests/module-manifest.test.ts
@@ -4,12 +4,13 @@ import {
   MODULE_LIFECYCLE_STAGES,
   packageRelativePathSchema,
   parseModuleManifest,
+  parsePermissionId,
   safeParseModuleManifest,
 } from "../src/index.js";
 
 function validManifest(): Record<string, unknown> {
   return {
-    manifestVersion: 2,
+    manifestVersion: 3,
     id: "calendar",
     name: "Calendar",
     version: "0.1.0",
@@ -18,7 +19,13 @@ function validManifest(): Record<string, unknown> {
       moduleSdk: "^0.1.0",
     },
     capabilities: [{ id: "calendar.events", version: "2.3.0" }],
-    permissions: ["calendar.events.read"],
+    permissions: [
+      {
+        id: "calendar.events.read",
+        baseline: "restricted",
+        description: "Read calendar events.",
+      },
+    ],
     storage: {
       buckets: [{ id: "assets", description: "Calendar-owned event assets." }],
     },
@@ -45,16 +52,23 @@ describe("module manifest contract", () => {
   it("creates isolated default collections for every parsed manifest", () => {
     const input = validManifest();
     delete input.storage;
+    delete input.permissions;
     const first = parseModuleManifest(input);
     const second = parseModuleManifest(input);
 
     first.dependencies.modules.required.people = "^0.1.0";
     first.dependencies.capabilities.required["people.records"] = "^1.0.0";
     first.storage.buckets.push({ id: "exports" });
+    first.permissions.push({
+      id: parsePermissionId("calendar.events.write"),
+      baseline: "restricted",
+      description: "Write calendar events.",
+    });
 
     expect(second.dependencies.modules.required).toEqual({});
     expect(second.dependencies.capabilities.required).toEqual({});
     expect(second.storage.buckets).toEqual([]);
+    expect(second.permissions).toEqual([]);
   });
 
   it("rejects unknown and presentation-oriented manifest properties", () => {
@@ -68,14 +82,16 @@ describe("module manifest contract", () => {
     const manifest = {
       ...validManifest(),
       capabilities: [{ id: "media.items", version: "1.0.0" }],
-      permissions: ["media.items.read"],
+      permissions: [
+        { id: "media.items.read", baseline: "restricted", description: "Read media items." },
+      ],
     };
     const result = safeParseModuleManifest(manifest);
 
     expect(result.success).toBe(false);
     if (!result.success) {
       expect(result.error.issues.map((issue) => issue.path.join("."))).toEqual(
-        expect.arrayContaining(["capabilities.0.id", "permissions.0"]),
+        expect.arrayContaining(["capabilities.0.id", "permissions.0.id"]),
       );
     }
   });
@@ -107,6 +123,32 @@ describe("module manifest contract", () => {
     }
   });
 
+  it("rejects a duplicate permission id at its exact issue path", () => {
+    const manifest = {
+      ...validManifest(),
+      permissions: [
+        {
+          id: "calendar.events.read",
+          baseline: "restricted",
+          description: "Read calendar events.",
+        },
+        {
+          id: "calendar.events.read",
+          baseline: "authenticated",
+          description: "Read calendar events, again.",
+        },
+      ],
+    };
+    const result = safeParseModuleManifest(manifest);
+
+    expect(result.success).toBe(false);
+    if (!result.success) {
+      expect(result.error.issues.map((issue) => issue.path.join("."))).toContain(
+        "permissions.1.id",
+      );
+    }
+  });
+
   it("rejects module and capability self-dependencies", () => {
     const manifest = {
       ...validManifest(),
@@ -151,6 +193,113 @@ describe("module manifest contract", () => {
     expect(safeParseModuleManifest({ ...validManifest(), manifestVersion: 1 }).success).toBe(false);
   });
 
+  it("rejects a version 2 manifest rather than silently interpreting it as version 3", () => {
+    // manifestVersion 2's own permissions shape (a bare string array) is also what version 3
+    // must reject — this manifest is invalid twice over, deliberately, since a real version 2
+    // manifest is exactly this shape.
+    const version2Manifest = {
+      ...validManifest(),
+      manifestVersion: 2,
+      permissions: ["calendar.events.read"],
+    };
+
+    expect(safeParseModuleManifest(version2Manifest).success).toBe(false);
+  });
+
+  it("rejects a bare permission-id string array instead of complete declarations", () => {
+    const manifest = { ...validManifest(), permissions: ["calendar.events.read"] };
+
+    expect(safeParseModuleManifest(manifest).success).toBe(false);
+  });
+
+  it("defaults to an empty permission list when omitted", () => {
+    const input = validManifest();
+    delete input.permissions;
+
+    const manifest = parseModuleManifest(input);
+    expect(manifest.permissions).toEqual([]);
+  });
+
+  it.each([
+    "public",
+    "authenticated",
+    "restricted",
+  ] as const)("accepts a %s permission declaration and preserves its exact baseline", (baseline) => {
+    const manifest = parseModuleManifest({
+      ...validManifest(),
+      permissions: [{ id: "calendar.events.read", baseline, description: "Read events." }],
+    });
+    expect(manifest.permissions).toEqual([
+      { id: "calendar.events.read", baseline, description: "Read events." },
+    ]);
+  });
+
+  it("rejects a permission declaration with a missing baseline", () => {
+    const manifest = {
+      ...validManifest(),
+      permissions: [{ id: "calendar.events.read", description: "Read events." }],
+    };
+    expect(safeParseModuleManifest(manifest).success).toBe(false);
+  });
+
+  it("rejects a permission declaration with an invalid baseline", () => {
+    const manifest = {
+      ...validManifest(),
+      permissions: [
+        { id: "calendar.events.read", baseline: "superuser", description: "Read events." },
+      ],
+    };
+    expect(safeParseModuleManifest(manifest).success).toBe(false);
+  });
+
+  it("rejects a permission declaration with a missing or blank description", () => {
+    const missingDescription = {
+      ...validManifest(),
+      permissions: [{ id: "calendar.events.read", baseline: "restricted" }],
+    };
+    const blankDescription = {
+      ...validManifest(),
+      permissions: [{ id: "calendar.events.read", baseline: "restricted", description: "   " }],
+    };
+    expect(safeParseModuleManifest(missingDescription).success).toBe(false);
+    expect(safeParseModuleManifest(blankDescription).success).toBe(false);
+  });
+
+  it("rejects a permission declaration with an oversized description", () => {
+    const manifest = {
+      ...validManifest(),
+      permissions: [
+        { id: "calendar.events.read", baseline: "restricted", description: "x".repeat(241) },
+      ],
+    };
+    expect(safeParseModuleManifest(manifest).success).toBe(false);
+  });
+
+  it("rejects a permission declaration with an unknown property", () => {
+    const manifest = {
+      ...validManifest(),
+      permissions: [
+        {
+          id: "calendar.events.read",
+          baseline: "restricted",
+          description: "Read events.",
+          scope: "calendar-event:123",
+        },
+      ],
+    };
+    expect(safeParseModuleManifest(manifest).success).toBe(false);
+  });
+
+  it("rejects a malformed permission id", () => {
+    const manifest = {
+      ...validManifest(),
+      permissions: [
+        { id: "Calendar.Events.Read", baseline: "restricted", description: "Read events." },
+      ],
+    };
+    expect(safeParseModuleManifest(manifest).success).toBe(false);
+  });
+
   it.each([
     "../outside.ts",
     "/absolute.ts",
diff --git a/templates/module/ekklesia.module.json b/templates/module/ekklesia.module.json
index ccfe877..6ea316a 100644
--- a/templates/module/ekklesia.module.json
+++ b/templates/module/ekklesia.module.json
@@ -1,5 +1,5 @@
 {
-  "manifestVersion": 2,
+  "manifestVersion": 3,
   "id": "example",
   "name": "Example Module",
   "description": "A minimal backend-only Ekklesia module template.",
@@ -14,7 +14,13 @@
       "version": "1.0.0"
     }
   ],
-  "permissions": ["example.records.read"],
+  "permissions": [
+    {
+      "id": "example.records.read",
+      "baseline": "restricted",
+      "description": "Read example records."
+    }
+  ],
   "storage": {
     "buckets": [
       {
diff --git a/templates/module/tests/manifest.test.ts b/templates/module/tests/manifest.test.ts
index 95ca89d..0b3b1e3 100644
--- a/templates/module/tests/manifest.test.ts
+++ b/templates/module/tests/manifest.test.ts
@@ -20,6 +20,13 @@ describe("module template", () => {
 
     expect(manifest.id).toBe("example");
     expect(manifest.capabilities).toEqual([{ id: "example.records", version: "1.0.0" }]);
+    expect(manifest.permissions).toEqual([
+      {
+        id: "example.records.read",
+        baseline: "restricted",
+        description: "Read example records.",
+      },
+    ]);
     expect(manifest.storage.buckets.map((bucket) => bucket.id)).toEqual(["records"]);
     expect(manifest.entrypoints.schema).toBe("./dist/schema/index.js");
     expect(manifest.entrypoints.migrations).toBe("./drizzle");