Wave 2 features: A2A live view (CEO chime-in + reply budget) and prompter memory (#297)

* feat(a2a): live view — watch fleet conversations, CEO chime-in, reply budget

A2A_MESSAGE_SENT published from A2AService.send (excerpt-capped) and
fanned through the existing /ws/system bridge; CEO-only admin REST for
conversations/messages + a reply route on the publish-bearing send path;
panel /a2a page with live transcript and a composer gated on task-linked
conversations. The matrix gains its one asymmetric rule: CEO may message
anyone, nobody may target the CEO — and agent replies inside a
CEO-opened conversation are hard-budgeted to one per CEO message
(per conversation, per agent), rejected with wait-don't-retry guidance.
Built subagent-driven (Sonnet 5), reviewed; v1 seams documented in the
map delta.

* feat(prompter): intake remembers the task history

Intake spawns now carry a per-project chronological digest of recent
tasks (capped: 15 lines/project, 4000 chars total — ~300-1000 tokens)
merged into the ambient layer, and the interviewer gets a bounded
search_past_tasks tool (one shared implementation behind the grok MCP
tool and the Claude SDK in-process tool) to check precedent
mid-conversation. Informational memory only — the sequencing analyzer
keeps ownership of ordering. Built subagent-driven (Sonnet 5), reviewed;
pre-existing conventions-ambient MegaTask-scope gap flagged, untouched.

---------

Co-authored-by: Renn F <rennf93@users.noreply.github.com>

panel/src/hooks/use-a2a-live.ts

@@ -0,0 +1,53 @@
+"use client";
+
+import { useMutation, useQuery, useQueryClient } from "@tanstack/react-query";
+import { a2aApi, type AdminReplyRequest } from "@/lib/api/a2a";
+
+export const a2aLiveKeys = {
+  all: ["a2a-live"] as const,
+  conversations: ["a2a-live", "conversations"] as const,
+  messages: (conversationId: string) =>
+    ["a2a-live", "messages", conversationId] as const,
+};
+
+// Conversation list — refreshed by WS `a2a.message` invalidation and the
+// manual Refresh button; a short staleTime keeps remounts reasonably fresh.
+export function useA2AConversations(limit?: number) {
+  return useQuery({
+    queryKey: [...a2aLiveKeys.conversations, limit ?? 50],
+    queryFn: () => a2aApi.listAdminConversations(limit),
+    staleTime: 30_000,
+  });
+}
+
+// Transcript for one conversation. WS frames for the selected conversation
+// invalidate this key; full bodies always come from REST (excerpts are capped).
+export function useA2AMessages(conversationId: string | null) {
+  return useQuery({
+    queryKey: a2aLiveKeys.messages(conversationId || ""),
+    queryFn: () => a2aApi.listAdminMessages(conversationId!),
+    enabled: !!conversationId,
+    staleTime: 30_000,
+  });
+}
+
+export interface ReplyAsCeoVariables extends AdminReplyRequest {
+  conversationId: string;
+}
+
+// CEO chime-in. The reply is a direct CEO->to_agent message (pairwise model);
+// invalidate the list (the CEO<->agent conversation appears/updates there) and
+// the watched transcript's messages.
+export function useReplyAsCeo() {
+  const queryClient = useQueryClient();
+  return useMutation({
+    mutationFn: ({ conversationId, ...reply }: ReplyAsCeoVariables) =>
+      a2aApi.replyAsCeo(conversationId, reply),
+    onSuccess: (_sent, variables) => {
+      queryClient.invalidateQueries({ queryKey: a2aLiveKeys.conversations });
+      queryClient.invalidateQueries({
+        queryKey: a2aLiveKeys.messages(variables.conversationId),
+      });
+    },
+  });
+}

panel/src/hooks/use-websocket.ts

@@ -44,6 +44,18 @@ export interface NotificationMessage {
   timestamp?: string;
 }
 
+export interface A2ASystemMessage {
+  type: "connected" | "a2a.message";
+  conversation_id?: string;
+  message_id?: string;
+  task_id?: string;
+  from_agent?: string;
+  to_agent?: string;
+  skill?: string | null;
+  body_excerpt?: string;
+  timestamp?: string;
+}
+
 export interface SessionMessage {
   type: "connected" | "message.new";
   message_id?: string;
@@ -296,6 +308,38 @@ export function useNotificationStream() {
   };
 }
 
+/**
+ * Subscribe to live A2A traffic on the operator stream (`/ws/system`).
+ *
+ * The bridge publishes an `a2a.message` frame for every persisted
+ * agent<->agent chat message. Frames carry a capped `body_excerpt` only —
+ * consumers invalidate their REST queries for full bodies (the A2A live view
+ * idiom), never render the excerpt as the message.
+ */
+export function useA2ALiveStream() {
+  const {
+    state,
+    lastMessage,
+    messages,
+    clearMessages,
+    isConnected,
+    isConnecting,
+  } = useWebSocket<A2ASystemMessage>("/system", undefined, true);
+
+  // Filter to A2A frames (the system stream also carries rate-limit/usage).
+  const a2aMessages = messages.filter((m) => m.type === "a2a.message");
+
+  return {
+    state,
+    lastMessage,
+    a2aMessages,
+    allMessages: messages,
+    clearMessages,
+    isConnected,
+    isConnecting,
+  };
+}
+
 // =============================================================================
 // Connection Status Hook (for UI indicators)
 // =============================================================================