Update files

packages/server/src/server/session.ts

@@ -68,6 +68,11 @@ import type {
 } from "./agent/agent-sdk-types.js";
 import { AgentStorage, type StoredAgentRecord } from "./agent/agent-storage.js";
 import { isValidAgentProvider, AGENT_PROVIDER_IDS } from "./agent/provider-manifest.js";
+import {
+  buildVoiceAgentMcpServerConfig,
+  buildVoiceModeSystemPrompt,
+  stripVoiceModeSystemPrompt,
+} from "./voice-config.js";
 import { isVoicePermissionAllowed } from "./voice-permission-policy.js";
 import {
   listDirectoryEntries,
@@ -130,18 +135,6 @@ const MAX_AGENTS_PER_PROJECT = 5;
  * Uses Claude Haiku for speed and cost efficiency.
  */
 const AUTO_GEN_MODEL = "haiku";
-const VOICE_AGENT_SYSTEM_INSTRUCTION = [
-  "You are the Paseo voice assistant.",
-  "The user cannot see your chat messages or tool calls.",
-  "Always use the speak tool for all user-facing communication.",
-  "Before calling any non-speak tool, first call speak with a short acknowledgement of what you heard and what you will do next.",
-  "For long-running work, use speak to provide progress updates before and during execution.",
-  "Treat the user input as transcribed speech and confirm intent in your first speak response to catch transcription errors early.",
-  "If the transcription seems incomplete, cut off, or ambiguous, ask a clarifying question via speak before taking action.",
-  "Use concise plain language suitable for speech output.",
-  "Never use bash, file-edit, or web tools directly.",
-  "Only use the paseo MCP tools.",
-].join(" ");
 
 function deriveRemoteProjectKey(remoteUrl: string | null): string | null {
   if (!remoteUrl) {
@@ -222,25 +215,6 @@ function escapeXmlText(value: string): string {
     .replace(/>/g, ">");
 }
 
-export function buildVoiceAgentMcpServerConfig(params: {
-  command: string;
-  baseArgs: string[];
-  socketPath: string;
-  env?: Record<string, string>;
-}): {
-  type: "stdio";
-  command: string;
-  args: string[];
-  env?: Record<string, string>;
-} {
-  return {
-    type: "stdio",
-    command: params.command,
-    args: [...params.baseArgs, "--socket", params.socketPath],
-    ...(params.env ? { env: params.env } : {}),
-  };
-}
-
 type ProcessingPhase = "idle" | "transcribing" | "llm";
 
 type NormalizedGitOptions = {
@@ -1541,13 +1515,6 @@ export class Session {
     return JSON.parse(JSON.stringify(servers)) as Record<string, McpServerConfig>;
   }
 
-  private buildVoiceModeSystemPrompt(existing?: string): string {
-    const chunks = [existing?.trim(), VOICE_AGENT_SYSTEM_INSTRUCTION].filter(
-      (value): value is string => Boolean(value && value.length > 0)
-    );
-    return chunks.join("\n\n");
-  }
-
   private buildVoiceModeMcpServers(
     existing: Record<string, McpServerConfig> | undefined,
     socketPath: string
@@ -1579,12 +1546,12 @@ export class Session {
     this.registerVoiceBridgeForAgent(agentId);
 
     const baseConfig: VoiceModeBaseConfig = {
-      systemPrompt: existing.config.systemPrompt,
+      systemPrompt: stripVoiceModeSystemPrompt(existing.config.systemPrompt),
       mcpServers: this.cloneMcpServers(existing.config.mcpServers),
     };
     this.voiceModeBaseConfig = baseConfig;
     const refreshOverrides: Partial<AgentSessionConfig> = {
-      systemPrompt: this.buildVoiceModeSystemPrompt(baseConfig.systemPrompt),
+      systemPrompt: buildVoiceModeSystemPrompt(baseConfig.systemPrompt, true),
       mcpServers: this.buildVoiceModeMcpServers(baseConfig.mcpServers, socketPath),
     };
 
@@ -1625,7 +1592,7 @@ export class Session {
       const baseConfig = this.voiceModeBaseConfig;
       try {
         await this.agentManager.refreshAgentFromPersistence(agentId, {
-          systemPrompt: baseConfig.systemPrompt,
+          systemPrompt: buildVoiceModeSystemPrompt(baseConfig.systemPrompt, false),
           mcpServers: this.cloneMcpServers(baseConfig.mcpServers),
         });
       } catch (error) {

packages/server/src/server/session.voice-mcp-config.test.ts

@@ -1,6 +1,10 @@
 import { describe, expect, test } from "vitest";
 
-import { buildVoiceAgentMcpServerConfig } from "./session.js";
+import {
+  buildVoiceAgentMcpServerConfig,
+  buildVoiceModeSystemPrompt,
+  stripVoiceModeSystemPrompt,
+} from "./voice-config.js";
 
 describe("voice MCP stdio config", () => {
   test("builds stdio MCP config for voice agent", () => {
@@ -25,3 +29,48 @@ describe("voice MCP stdio config", () => {
     });
   });
 });
+
+describe("voice mode prompt instructions", () => {
+  test("builds enabled voice instructions and preserves base prompt", () => {
+    const prompt = buildVoiceModeSystemPrompt("Base system prompt", true);
+
+    expect(prompt).toContain("Base system prompt");
+    expect(prompt).toContain("<paseo_voice_mode>");
+    expect(prompt).toContain("Paseo voice mode is now on.");
+    expect(prompt).toContain("Always use the speak tool for all user-facing communication.");
+    expect(prompt).toContain("</paseo_voice_mode>");
+  });
+
+  test("builds disabled voice instructions and supersedes previous voice block", () => {
+    const existing = [
+      "Base system prompt",
+      "<paseo_voice_mode>",
+      "legacy voice instruction",
+      "</paseo_voice_mode>",
+    ].join("\n\n");
+
+    const prompt = buildVoiceModeSystemPrompt(existing, false);
+
+    expect(prompt).toContain("Base system prompt");
+    expect(prompt).toContain("Paseo voice mode is now off.");
+    expect(prompt).toContain("Ignore any earlier Paseo voice mode instructions in this thread.");
+    expect(prompt.match(/<paseo_voice_mode>/g)?.length ?? 0).toBe(1);
+    expect(prompt).not.toContain("legacy voice instruction");
+  });
+
+  test("strips voice blocks from persisted prompt", () => {
+    const existing = [
+      "Base system prompt",
+      "<paseo_voice_mode>",
+      "legacy voice instruction",
+      "</paseo_voice_mode>",
+    ].join("\n\n");
+
+    expect(stripVoiceModeSystemPrompt(existing)).toBe("Base system prompt");
+    expect(
+      stripVoiceModeSystemPrompt(
+        ["<paseo_voice_mode>", "legacy voice instruction", "</paseo_voice_mode>"].join("\n\n")
+      )
+    ).toBeUndefined();
+  });
+});

packages/server/src/server/voice-config.ts

@@ -0,0 +1,75 @@
+const VOICE_PROMPT_BLOCK_START = "<paseo_voice_mode>";
+const VOICE_PROMPT_BLOCK_END = "</paseo_voice_mode>";
+
+const VOICE_AGENT_SYSTEM_INSTRUCTION = [
+  "Paseo voice mode is now on.",
+  "You are the Paseo voice assistant.",
+  "The user cannot see your chat messages or tool calls.",
+  "Always use the speak tool for all user-facing communication.",
+  "Before calling any non-speak tool, first call speak with a short acknowledgement of what you heard and what you will do next.",
+  "For long-running work, use speak to provide progress updates before and during execution.",
+  "Treat the user input as transcribed speech.",
+  "If the user intent is clear, proceed without extra confirmation.",
+  "If the transcription seems incomplete, cut off, ambiguous, or may contain a non-obvious mistake or misspelling, ask a clarifying question via speak before taking action.",
+  "Use concise plain language suitable for speech output.",
+].join(" ");
+
+const VOICE_AGENT_DISABLED_INSTRUCTION = [
+  "Paseo voice mode is now off.",
+  "Ignore any earlier Paseo voice mode instructions in this thread.",
+].join(" ");
+
+function escapeRegExp(value: string): string {
+  return value.replace(/[.*+?^${}()|[\]\\]/g, "\\$&");
+}
+
+function makeVoicePromptBlockRegex(): RegExp {
+  return new RegExp(
+    `${escapeRegExp(VOICE_PROMPT_BLOCK_START)}[\\s\\S]*?${escapeRegExp(
+      VOICE_PROMPT_BLOCK_END
+    )}`,
+    "g"
+  );
+}
+
+export function stripVoiceModeSystemPrompt(existing?: string): string | undefined {
+  const trimmed = existing?.trim();
+  if (!trimmed) {
+    return undefined;
+  }
+  const stripped = trimmed.replace(makeVoicePromptBlockRegex(), "").trim();
+  return stripped.length > 0 ? stripped : undefined;
+}
+
+export function buildVoiceModeSystemPrompt(existing: string | undefined, enabled: boolean): string {
+  const basePrompt = stripVoiceModeSystemPrompt(existing);
+  const voiceInstruction = enabled
+    ? VOICE_AGENT_SYSTEM_INSTRUCTION
+    : VOICE_AGENT_DISABLED_INSTRUCTION;
+  const voiceBlock = [VOICE_PROMPT_BLOCK_START, voiceInstruction, VOICE_PROMPT_BLOCK_END].join(
+    "\n"
+  );
+
+  return [basePrompt, voiceBlock]
+    .filter((entry): entry is string => typeof entry === "string" && entry.length > 0)
+    .join("\n\n");
+}
+
+export function buildVoiceAgentMcpServerConfig(params: {
+  command: string;
+  baseArgs: string[];
+  socketPath: string;
+  env?: Record<string, string>;
+}): {
+  type: "stdio";
+  command: string;
+  args: string[];
+  env?: Record<string, string>;
+} {
+  return {
+    type: "stdio",
+    command: params.command,
+    args: [...params.baseArgs, "--socket", params.socketPath],
+    ...(params.env ? { env: params.env } : {}),
+  };
+}