chat: inject an Officer system prompt telling OpenCode its working directory (step 2)

OpenCode can't set a real per-session cwd (every session runs in the fixed server's
dir), so we tell the model its working directory via a system prompt appended to
OpenCode's own — sent as a system message, so it never appears in the visible chat
(verified against source + live). Claude doesn't need this (it honors cwd natively).

- client.postMessage(…, system?) forwards a `system` string on the message.
- send-opencode builds the Officer prompt from `workingDir` and sends it every turn.
- websocket: workingDir = the resolved cwd, except the general /chat (whose cwd is the
  claude_sessions grouping placeholder) uses the user's home.

Verified live: with the prompt, the model reports the injected dir as its cwd.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

src/servers/api/chat/opencode/client.ts

@@ -117,11 +117,21 @@ class ServerConnection {
     return session.id;
   }
 
-  async postMessage(sessionId: string, providerID: string, modelID: string, text: string): Promise<void> {
-    await this.postJson(`/session/${sessionId}/message`, {
+  async postMessage(
+    sessionId: string,
+    providerID: string,
+    modelID: string,
+    text: string,
+    system?: string,
+  ): Promise<void> {
+    // `system` is appended to OpenCode's built-in system prompt (additive, not an override) and is
+    // sent to the LLM as a system message — it never appears as a visible chat part.
+    const body: Record<string, unknown> = {
       model: { providerID, modelID },
       parts: [{ type: 'text', text }],
-    });
+    };
+    if (system) body.system = system;
+    await this.postJson(`/session/${sessionId}/message`, body);
   }
 
   async abort(sessionId: string): Promise<void> {