fix: preserve structured ACP prompt blocks

src/cli-core.ts

@@ -38,6 +38,7 @@ import {
   type NormalizedOutputError,
 } from "./error-normalization.js";
 import { flushPerfMetricsCapture, installPerfMetricsCapture } from "./perf-metrics-capture.js";
+import { mergePromptSourceWithText, parsePromptSource, textPrompt } from "./prompt-content.js";
 import { runQueueOwnerFromEnv } from "./queue-owner-env.js";
 import {
   DEFAULT_HISTORY_LIMIT,
@@ -88,25 +89,22 @@ async function readPrompt(
   promptParts: string[],
   filePath: string | undefined,
   cwd: string,
-): Promise<string> {
+): Promise<import("./types.js").PromptInput> {
   if (filePath) {
     const source =
       filePath === "-"
         ? await readPromptInputFromStdin()
         : await fs.readFile(path.resolve(cwd, filePath), "utf8");
-    const pieces = [source.trim(), promptParts.join(" ").trim()].filter(
-      (value) => value.length > 0,
-    );
-    const prompt = pieces.join("\n\n").trim();
-    if (!prompt) {
+    const prompt = mergePromptSourceWithText(source, promptParts.join(" "));
+    if (prompt.length === 0) {
       throw new InvalidArgumentError("Prompt from --file is empty");
     }
     return prompt;
   }
 
   const joined = promptParts.join(" ").trim();
   if (joined.length > 0) {
-    return joined;
+    return textPrompt(joined);
   }
 
   if (process.stdin.isTTY) {
@@ -115,8 +113,8 @@ async function readPrompt(
     );
   }
 
-  const prompt = (await readPromptInputFromStdin()).trim();
-  if (!prompt) {
+  const prompt = parsePromptSource(await readPromptInputFromStdin());
+  if (prompt.length === 0) {
     throw new InvalidArgumentError("Prompt from stdin is empty");
   }
 
@@ -251,7 +249,7 @@ async function handlePrompt(
   await printPromptSessionBanner(record, agent.cwd, outputPolicy.format, outputPolicy.jsonStrict);
   const result = await sendSession({
     sessionId: record.acpxRecordId,
-    message: prompt,
+    prompt,
     mcpServers: config.mcpServers,
     permissionMode,
     nonInteractivePermissions: globalFlags.nonInteractivePermissions,
@@ -327,7 +325,7 @@ async function handleExec(
   const result = await runOnce({
     agentCommand: agent.agentCommand,
     cwd: agent.cwd,
-    message: prompt,
+    prompt,
     mcpServers: config.mcpServers,
     permissionMode,
     nonInteractivePermissions: globalFlags.nonInteractivePermissions,

src/client.ts

@@ -40,6 +40,7 @@ import {
 } from "./errors.js";
 import { FileSystemHandlers } from "./filesystem.js";
 import { classifyPermissionDecision, resolvePermissionRequest } from "./permissions.js";
+import { textPrompt } from "./prompt-content.js";
 import { extractRuntimeSessionId } from "./runtime-session-id.js";
 import { TimeoutError, withTimeout } from "./session-runtime-helpers.js";
 import { TerminalManager } from "./terminal.js";
@@ -48,6 +49,7 @@ import type {
   NonInteractivePermissionPolicy,
   PermissionMode,
   PermissionStats,
+  PromptInput,
 } from "./types.js";
 
 type CommandParts = {
@@ -976,7 +978,7 @@ export class AcpClient {
     };
   }
 
-  async prompt(sessionId: string, text: string): Promise<PromptResponse> {
+  async prompt(sessionId: string, prompt: PromptInput | string): Promise<PromptResponse> {
     const connection = this.getConnection();
     const restoreConsoleError = this.options.suppressSdkConsoleErrors
       ? installSdkConsoleErrorSuppression()
@@ -986,12 +988,7 @@ export class AcpClient {
     try {
       promptPromise = connection.prompt({
         sessionId,
-        prompt: [
-          {
-            type: "text",
-            text,
-          },
-        ],
+        prompt: typeof prompt === "string" ? textPrompt(prompt) : prompt,
       });
     } catch (error) {
       restoreConsoleError?.();

src/prompt-content.ts

@@ -0,0 +1,130 @@
+import type { ContentBlock } from "@agentclientprotocol/sdk";
+
+export type PromptInput = ContentBlock[];
+
+function asRecord(value: unknown): Record<string, unknown> | undefined {
+  if (!value || typeof value !== "object" || Array.isArray(value)) {
+    return undefined;
+  }
+  return value as Record<string, unknown>;
+}
+
+function isTextBlock(value: unknown): value is Extract<ContentBlock, { type: "text" }> {
+  const record = asRecord(value);
+  return record?.type === "text" && typeof record.text === "string";
+}
+
+function isImageBlock(value: unknown): value is Extract<ContentBlock, { type: "image" }> {
+  const record = asRecord(value);
+  return (
+    record?.type === "image" &&
+    typeof record.mimeType === "string" &&
+    typeof record.data === "string"
+  );
+}
+
+function isResourceLinkBlock(
+  value: unknown,
+): value is Extract<ContentBlock, { type: "resource_link" }> {
+  const record = asRecord(value);
+  return (
+    record?.type === "resource_link" &&
+    typeof record.uri === "string" &&
+    (record.title === undefined || typeof record.title === "string") &&
+    (record.name === undefined || typeof record.name === "string")
+  );
+}
+
+function isResourcePayload(value: unknown): boolean {
+  const record = asRecord(value);
+  if (!record || typeof record.uri !== "string") {
+    return false;
+  }
+  return record.text === undefined || typeof record.text === "string";
+}
+
+function isResourceBlock(value: unknown): value is Extract<ContentBlock, { type: "resource" }> {
+  const record = asRecord(value);
+  return record?.type === "resource" && isResourcePayload(record.resource);
+}
+
+function isContentBlock(value: unknown): value is ContentBlock {
+  return (
+    isTextBlock(value) ||
+    isImageBlock(value) ||
+    isResourceLinkBlock(value) ||
+    isResourceBlock(value)
+  );
+}
+
+export function isPromptInput(value: unknown): value is PromptInput {
+  return Array.isArray(value) && value.every((entry) => isContentBlock(entry));
+}
+
+export function textPrompt(text: string): PromptInput {
+  return [
+    {
+      type: "text",
+      text,
+    },
+  ];
+}
+
+function parseStructuredPrompt(source: string): PromptInput | undefined {
+  if (!source.startsWith("[")) {
+    return undefined;
+  }
+  try {
+    const parsed = JSON.parse(source) as unknown;
+    return isPromptInput(parsed) ? parsed : undefined;
+  } catch {
+    return undefined;
+  }
+}
+
+export function parsePromptSource(source: string): PromptInput {
+  const trimmed = source.trim();
+  const structured = parseStructuredPrompt(trimmed);
+  if (structured) {
+    return structured;
+  }
+  if (!trimmed) {
+    return [];
+  }
+  return textPrompt(trimmed);
+}
+
+export function mergePromptSourceWithText(source: string, suffixText: string): PromptInput {
+  const prompt = parsePromptSource(source);
+  const appended = suffixText.trim();
+  if (!appended) {
+    return prompt;
+  }
+  if (prompt.length === 0) {
+    return textPrompt(appended);
+  }
+  return [...prompt, ...textPrompt(appended)];
+}
+
+export function promptToDisplayText(prompt: PromptInput): string {
+  return prompt
+    .map((block) => {
+      switch (block.type) {
+        case "text":
+          return block.text;
+        case "resource_link":
+          return block.title ?? block.name ?? block.uri;
+        case "resource":
+          return "text" in block.resource && typeof block.resource.text === "string"
+            ? block.resource.text
+            : block.resource.uri;
+        case "image":
+          return `[image] ${block.mimeType}`;
+        default:
+          return "";
+      }
+    })
+    .filter((entry) => entry.trim().length > 0)
+    .join("\n\n")
+    .trim();
+}

src/queue-ipc-server.ts

@@ -2,12 +2,13 @@ import net from "node:net";
 import type { SetSessionConfigOptionResponse } from "@agentclientprotocol/sdk";
 import { normalizeOutputError } from "./error-normalization.js";
 import { recordPerfDuration } from "./perf-metrics.js";
+import { textPrompt } from "./prompt-content.js";
 import {
   parseQueueRequest,
   type QueueOwnerErrorMessage,
   type QueueOwnerMessage,
 } from "./queue-messages.js";
-import type { NonInteractivePermissionPolicy, PermissionMode } from "./types.js";
+import type { NonInteractivePermissionPolicy, PermissionMode, PromptInput } from "./types.js";
 
 type QueueOwnerSocketLease = {
   socketPath: string;
@@ -71,6 +72,7 @@ function writeQueueMessage(socket: net.Socket, message: QueueOwnerMessage): void
 export type QueueTask = {
   requestId: string;
   message: string;
+  prompt: PromptInput;
   permissionMode: PermissionMode;
   nonInteractivePermissions?: NonInteractivePermissionPolicy;
   timeoutMs?: number;
@@ -440,6 +442,7 @@ export class SessionQueueOwner {
       const task: QueueTask = {
         requestId: request.requestId,
         message: request.message,
+        prompt: request.prompt ?? textPrompt(request.message),
         permissionMode: request.permissionMode,
         nonInteractivePermissions: request.nonInteractivePermissions,
         timeoutMs: request.timeoutMs,

src/queue-ipc.ts

@@ -32,6 +32,7 @@ import type {
   OutputErrorEmissionPolicy,
   OutputFormatter,
   PermissionMode,
+  PromptInput,
   SessionEnqueueResult,
   SessionSendOutcome,
 } from "./types.js";
@@ -220,6 +221,7 @@ function assertOwnerGeneration(
 export type SubmitToQueueOwnerOptions = {
   sessionId: string;
   message: string;
+  prompt?: PromptInput;
   permissionMode: PermissionMode;
   nonInteractivePermissions?: NonInteractivePermissionPolicy;
   outputFormatter: OutputFormatter;
@@ -246,6 +248,7 @@ async function submitToQueueOwner(
     requestId,
     ownerGeneration: owner.ownerGeneration,
     message: options.message,
+    prompt: options.prompt,
     permissionMode: options.permissionMode,
     nonInteractivePermissions: options.nonInteractivePermissions,
     timeoutMs: options.timeoutMs,

src/queue-messages.ts

@@ -1,5 +1,6 @@
 import type { SetSessionConfigOptionResponse } from "@agentclientprotocol/sdk";
 import { isAcpJsonRpcMessage } from "./acp-jsonrpc.js";
+import { isPromptInput, textPrompt } from "./prompt-content.js";
 import {
   OUTPUT_ERROR_CODES,
   OUTPUT_ERROR_ORIGINS,
@@ -11,6 +12,7 @@ import type {
   AcpJsonRpcMessage,
   NonInteractivePermissionPolicy,
   PermissionMode,
+  PromptInput,
   SessionSendResult,
 } from "./types.js";
 
@@ -19,6 +21,7 @@ export type QueueSubmitRequest = {
   requestId: string;
   ownerGeneration?: number;
   message: string;
+  prompt?: PromptInput;
   permissionMode: PermissionMode;
   nonInteractivePermissions?: NonInteractivePermissionPolicy;
   timeoutMs?: number;
@@ -204,9 +207,12 @@ export function parseQueueRequest(raw: unknown): QueueRequest | null {
           ? request.suppressSdkConsoleErrors
           : null;
 
+    const prompt =
+      request.prompt == null ? undefined : isPromptInput(request.prompt) ? request.prompt : null;
     if (
       typeof request.message !== "string" ||
       !isPermissionMode(request.permissionMode) ||
+      prompt === null ||
       nonInteractivePermissions === null ||
       suppressSdkConsoleErrors === null ||
       typeof request.waitForCompletion !== "boolean"
@@ -219,6 +225,7 @@ export function parseQueueRequest(raw: unknown): QueueRequest | null {
       requestId: request.requestId,
       ownerGeneration,
       message: request.message,
+      prompt: prompt ?? textPrompt(request.message),
       permissionMode: request.permissionMode,
       nonInteractivePermissions,
       timeoutMs,

src/session-conversation-model.ts

@@ -7,8 +7,10 @@ import type {
   ToolCallUpdate,
   UsageUpdate,
 } from "@agentclientprotocol/sdk";
+import { textPrompt } from "./prompt-content.js";
 import type {
   ClientOperation,
+  PromptInput,
   SessionAcpxState,
   SessionConversation,
   SessionAgentContent,
@@ -489,18 +491,28 @@ export function appendLegacyHistory(
 
 export function recordPromptSubmission(
   conversation: SessionConversation,
-  prompt: string,
+  prompt: PromptInput | string,
   timestamp = isoNow(),
 ): void {
-  const text = prompt.trim();
-  if (!text) {
+  const normalizedPrompt = typeof prompt === "string" ? textPrompt(prompt) : prompt;
+  const userContent = normalizedPrompt
+    .map((content) => contentToUserContent(content))
+    .filter((content) => content !== undefined);
+  if (userContent.length === 0) {
     return;
   }
 
   conversation.messages.push({
     User: {
       id: nextUserMessageId(),
-      content: [{ Text: trimRuntimeText(text, MAX_RUNTIME_AGENT_TEXT_CHARS) }],
+      content: userContent.map((content) => {
+        if ("Text" in content) {
+          return {
+            Text: trimRuntimeText(content.Text, MAX_RUNTIME_AGENT_TEXT_CHARS),
+          };
+        }
+        return content;
+      }),
     },
   });
   updateConversationTimestamp(conversation, timestamp);