fix(ai): append system prompt to codex bridge message instead of converting to input

Previously the system prompt was converted to an input message in convertMessages,
then stripped out by filterPiSystemPrompts. Now the system prompt is passed directly
to transformRequestBody and appended after CODEX_PI_BRIDGE in the bridge message.

packages/ai/src/providers/openai-codex-responses.ts

@@ -123,6 +123,7 @@ export const streamOpenAICodexResponses: StreamFunction<"openai-codex-responses"
 				codexInstructions,
 				codexOptions,
 				options?.codexMode ?? true,
+				context.systemPrompt,
 			);
 
 			const headers = createCodexHeaders(model.headers, accountId, apiKey, transformedBody.prompt_cache_key);
@@ -477,14 +478,6 @@ function convertMessages(model: Model<"openai-codex-responses">, context: Contex
 
 	const transformedMessages = transformMessages(context.messages, model);
 
-	if (context.systemPrompt) {
-		const role = model.reasoning ? "developer" : "system";
-		messages.push({
-			role,
-			content: sanitizeSurrogates(context.systemPrompt),
-		});
-	}
-
 	let msgIndex = 0;
 	for (const msg of transformedMessages) {
 		if (msg.role === "user") {