refactor(permissions): preserve pre-PR behavior in three spots

PR #5 review flagged three behavior changes that shouldn't have slipped
in. This commit reverts each to match the pre-refactor behavior exactly.

1. User upsert ordering. Split the router hook into two setters:
   setSenderResolver (runs before agent resolution) and setAccessGate
   (runs after). Restores the pre-PR sequence where the users row is
   upserted even if the message is dropped by wiring or trigger rules.

2. dropped_messages audit. Moved src/modules/permissions/db/dropped-messages.ts
   back to src/db/dropped-messages.ts. The table is core audit infra, not
   permissions-specific. Router re-writes rows for no_agent_wired and
   no_trigger_match; the access gate writes rows for policy refusals.

3. Permissionless container fallback. Dropped. poll-loop restores the
   original deny-all check when NANOCLAW_ADMIN_USER_IDS is empty.

Module contract doc updated with the two-hook shape.

Validation: host build clean, 137/137 host tests, 17/17 container
tests, typecheck clean, service boots to "NanoClaw running" with
permissions module registering both hooks and clean SIGTERM shutdown.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

src/modules/permissions/db/dropped-messages.ts

@@ -1,44 +0,0 @@
-import { getDb } from '../../../db/connection.js';
-
-export interface UnregisteredSender {
-  channel_type: string;
-  platform_id: string;
-  user_id: string | null;
-  sender_name: string | null;
-  reason: string;
-  messaging_group_id: string | null;
-  agent_group_id: string | null;
-  message_count: number;
-  first_seen: string;
-  last_seen: string;
-}
-
-export function recordDroppedMessage(msg: {
-  channel_type: string;
-  platform_id: string;
-  user_id: string | null;
-  sender_name: string | null;
-  reason: string;
-  messaging_group_id: string | null;
-  agent_group_id: string | null;
-}): void {
-  const now = new Date().toISOString();
-  getDb()
-    .prepare(
-      `INSERT INTO unregistered_senders (channel_type, platform_id, user_id, sender_name, reason, messaging_group_id, agent_group_id, message_count, first_seen, last_seen)
-       VALUES (@channel_type, @platform_id, @user_id, @sender_name, @reason, @messaging_group_id, @agent_group_id, 1, @now, @now)
-       ON CONFLICT (channel_type, platform_id) DO UPDATE SET
-         user_id = COALESCE(excluded.user_id, unregistered_senders.user_id),
-         sender_name = COALESCE(excluded.sender_name, unregistered_senders.sender_name),
-         reason = excluded.reason,
-         message_count = unregistered_senders.message_count + 1,
-         last_seen = excluded.last_seen`,
-    )
-    .run({ ...msg, now });
-}
-
-export function getUnregisteredSenders(limit = 50): UnregisteredSender[] {
-  return getDb()
-    .prepare('SELECT * FROM unregistered_senders ORDER BY last_seen DESC LIMIT ?')
-    .all(limit) as UnregisteredSender[];
-}

src/modules/permissions/index.ts

@@ -1,25 +1,25 @@
 /**
  * Permissions module — sender resolution + access gate.
  *
- * Registers a single inbound-gate via setInboundGate(). The gate owns:
- *   1. Sender resolution: parse the channel adapter's payload, derive a
- *      namespaced user id, and upsert the `users` row on first sight so
- *      role/access lookups land on a real record thereafter.
- *   2. Access decision: owners → global admins → scoped admins → members.
- *   3. Unknown-sender policy: strict drops; request_approval is a TODO
- *      (pending the `add_group_member` action kind).
- *   4. Audit trail: drops get logged into `dropped_messages`.
+ * Registers two hooks into the core router:
+ *   1. setSenderResolver — runs before agent resolution. Parses the payload,
+ *      derives a namespaced user id, and upserts the `users` row on first
+ *      sight. Returns null when the payload doesn't carry enough to identify
+ *      a sender.
+ *   2. setAccessGate — runs after agent resolution. Enforces the
+ *      unknown_sender_policy (strict/request_approval/public) and the
+ *      owner/global-admin/scoped-admin/member access hierarchy. Records its
+ *      own `dropped_messages` row on refusal (structural drops are recorded
+ *      by core).
  *
- * Without this module: core's router defaults to allow-all (PR #2), every
- * message routes through, and no users table is needed. Drops are not
- * recorded anywhere. Admin commands inside the container fall back to
- * permissionless mode (see formatter.ts).
+ * Without this module: sender resolution is a no-op (userId=null); the
+ * access gate is not registered and core defaults to allow-all.
  */
-import { setInboundGate, type InboundEvent, type InboundGateResult } from '../../router.js';
+import { recordDroppedMessage } from '../../db/dropped-messages.js';
+import { setAccessGate, setSenderResolver, type AccessGateResult, type InboundEvent } from '../../router.js';
 import { log } from '../../log.js';
 import type { MessagingGroup } from '../../types.js';
 import { canAccessAgentGroup } from './access.js';
-import { recordDroppedMessage } from './db/dropped-messages.js';
 import { getUser, upsertUser } from './db/users.js';
 
 function extractAndUpsertUser(event: InboundEvent): string | null {
@@ -111,24 +111,24 @@ function handleUnknownSender(
   // 'public' should have been handled before the gate; fall through silently.
 }
 
-setInboundGate((event, mg, agentGroupId): InboundGateResult => {
-  const userId = extractAndUpsertUser(event);
+setSenderResolver(extractAndUpsertUser);
 
+setAccessGate((event, userId, mg, agentGroupId): AccessGateResult => {
   // Public channels skip the access check entirely.
   if (mg.unknown_sender_policy === 'public') {
-    return { allowed: true, userId };
+    return { allowed: true };
   }
 
   if (!userId) {
     handleUnknownSender(mg, null, agentGroupId, 'unknown_user', event);
-    return { allowed: false, userId: null, reason: 'unknown_user' };
+    return { allowed: false, reason: 'unknown_user' };
   }
 
   const decision = canAccessAgentGroup(userId, agentGroupId);
   if (decision.allowed) {
-    return { allowed: true, userId };
+    return { allowed: true };
   }
 
   handleUnknownSender(mg, userId, agentGroupId, decision.reason, event);
-  return { allowed: false, userId, reason: decision.reason };
+  return { allowed: false, reason: decision.reason };
 });