fix(server): Codex native subagents inherit parent options

apps/server/src/mcp/toolkits/orchestrator/tools.test.ts

@@ -14,3 +14,20 @@ it("states t3_thread_read page size limits in the parameter descriptions", () =>
     expect(schema.properties[name]?.description, name).toContain(maximum);
   }
 });
+
+// Codex native spawn copies the parent's reasoning effort and service tier, so inherited
+// Codex options alone must not push plain subagent requests onto delegate_task (#23525).
+it.each(["delegate_task", "create_threads"] as const)(
+  "%s does not route every inherited-options subagent away from native spawn",
+  (name) => {
+    expect(OrchestratorToolkit.tools[name].description).not.toMatch(
+      /no model options are requested or inherited/i,
+    );
+  },
+);
+
+it("delegate_task says inherited Codex options alone do not require it", () => {
+  expect(OrchestratorToolkit.tools.delegate_task.description).toMatch(
+    /inherited Codex options alone do not require delegate_task/i,
+  );
+});

apps/server/src/mcp/toolkits/orchestrator/tools.ts

@@ -67,7 +67,7 @@ const OrchestratorCapabilitiesTool = Tool.make("orchestrator_capabilities", {
 
 const DelegateTaskTool = Tool.make("delegate_task", {
   description:
-    "Needs an agent running inside a T3 thread. Delegate one task to a T3-owned child agent/subagent of THIS thread and run it with only the supplied task prompt, without copying parent conversation history. Choose providers and models from orchestrator_capabilities, which uses the same live catalog as the composer. Prefer native subagent tools for same-provider work only when they support the chosen model and no model options are requested or inherited. Use this for any model missing from the native tool, including same-provider work, for cross-provider work, or for explicitly T3-owned child tasks. For every T3 delegated review round, call delegate_task again with the original brief, prior findings, responses, and unresolved objections in the task prompt. Track each round by its own taskId and use a distinct clientRequestId per round, stable across retries of that round. The childThreadId is backing storage, not the target for starting another delegated review round through t3_thread_send. Set isolation to worktree for a separate checkout based on the parent local commit, with project setup completed before the child starts. Uncommitted edits are not copied; omitted isolation shares the parent checkout. Use delegate_task when model options are requested or the child keeps the parent's provider instance and model with nonempty inheritedOptions from orchestrator_capabilities, even for native-supported models. Native subagents cannot reliably apply T3 model options. Set target.options using advertised IDs: Claude {effort: 'max', fastMode: true}; Codex {reasoningEffort: 'max', serviceTier: 'priority'}. Check providerInstanceId, model, and options in the result. Provider/model and modes inherit unless overridden. Options inherit only for the same instance and model; explicit entries override individual values and [] or {} clears them. A requested runtimeMode must be one the target provider lists in orchestrator_capabilities runtimeModes; an inherited one steps down to the broadest mode it offers. Leave runtimeMode unset unless the user asks for a supervised child: approval-required and auto-accept-edits usually make the child wait for a person at each action that needs approval. Prefer mode='async' for long work; mode='wait' blocks until completion, a blocking question, a pending approval, or timeout. If hasPendingQuestion, pass threadId=childThreadId to t3_pending_request_list, t3_pending_request_read (with requestId) and t3_pending_request_respond, or ask the user; the task's completion still wakes this thread. If pendingApproval is set, read the fresh full request in childThreadId before responding: the record can outlive its provider callback. It grants no approval authority; decisions remain with the child's existing approval holder, and pending request tools answer questions only. The child keeps running and its completion still wakes this thread. timeoutMs on mode=wait is only the parent's wait budget and does not cancel the child. waitTimedOut on that wait call means the timeout fired; keep that taskId and read status on later task_status. An async child's completion wakes this thread through a notification, steered into active turns where supported or queued otherwise, so end the turn instead of polling or spawning watchers; use task_status only when the result is needed mid-turn.",
+    "Needs an agent running inside a T3 thread. Delegate one task to a T3-owned child agent/subagent of THIS thread and run it with only the supplied task prompt, without copying parent conversation history. Choose providers and models from orchestrator_capabilities, which uses the same live catalog as the composer. Prefer native subagent tools for same-provider work only when they support the chosen model and carry every requested or inherited model option. Use this for any model missing from the native tool, including same-provider work, for cross-provider work, or for explicitly T3-owned child tasks. For every T3 delegated review round, call delegate_task again with the original brief, prior findings, responses, and unresolved objections in the task prompt. Track each round by its own taskId and use a distinct clientRequestId per round, stable across retries of that round. The childThreadId is backing storage, not the target for starting another delegated review round through t3_thread_send. Set isolation to worktree for a separate checkout based on the parent local commit, with project setup completed before the child starts. Uncommitted edits are not copied; omitted isolation shares the parent checkout. Codex native subagents keep the parent's reasoning effort and supported service tier unless a role or configured subagent default overrides them, so inherited Codex options alone do not require delegate_task. Use delegate_task even for native-supported models when model options are requested, when a role or subagent default would change inherited settings, or when the child keeps the parent's provider instance and model with nonempty inheritedOptions from orchestrator_capabilities on any other provider. Other native subagents cannot reliably apply T3 model options such as Claude effort or fastMode. Set target.options using advertised IDs: Claude {effort: 'max', fastMode: true}; Codex {reasoningEffort: 'max', serviceTier: 'priority'}. Check providerInstanceId, model, and options in the result. Provider/model and modes inherit unless overridden. Options inherit only for the same instance and model; explicit entries override individual values and [] or {} clears them. A requested runtimeMode must be one the target provider lists in orchestrator_capabilities runtimeModes; an inherited one steps down to the broadest mode it offers. Leave runtimeMode unset unless the user asks for a supervised child: approval-required and auto-accept-edits usually make the child wait for a person at each action that needs approval. Prefer mode='async' for long work; mode='wait' blocks until completion, a blocking question, a pending approval, or timeout. If hasPendingQuestion, pass threadId=childThreadId to t3_pending_request_list, t3_pending_request_read (with requestId) and t3_pending_request_respond, or ask the user; the task's completion still wakes this thread. If pendingApproval is set, read the fresh full request in childThreadId before responding: the record can outlive its provider callback. It grants no approval authority; decisions remain with the child's existing approval holder, and pending request tools answer questions only. The child keeps running and its completion still wakes this thread. timeoutMs on mode=wait is only the parent's wait budget and does not cancel the child. waitTimedOut on that wait call means the timeout fired; keep that taskId and read status on later task_status. An async child's completion wakes this thread through a notification, steered into active turns where supported or queued otherwise, so end the turn instead of polling or spawning watchers; use task_status only when the result is needed mid-turn.",
   parameters: OrchestratorMcpDelegateTaskInput,
   success: OrchestratorMcpDelegateTaskResult,
   failure: OrchestratorMcpFailure,
@@ -181,7 +181,7 @@ const RequestSecretTool = Tool.make("request_secret", {
 
 const CreateThreadsTool = Tool.make("create_threads", {
   description:
-    "Needs an agent running inside a T3 thread. Create one or more ORDINARY TOP-LEVEL T3 conversations. This is not delegation and does not create child agents/subagents. For delegated work, choose models from orchestrator_capabilities. Prefer native subagents only when they support the chosen model and no model options are requested or inherited; otherwise call delegate_task, including for same-provider work. This includes keeping the parent's provider instance and model with nonempty inheritedOptions from orchestrator_capabilities. Use create_threads for a batch of separate top-level threads sharing this checkout. Prefer t3_thread_launch for a single thread. Both require the user to request separate/new/top-level threads or conversations. Each entry may override provider, model, options, runtime mode, and interaction mode; omitted settings inherit. Project, branch, and worktree always inherit and cannot be overridden here. For independent implementation or a PR stack in its own worktree, use t3_thread_launch with workspaceStrategy instead of asking the agent to create a worktree in its prompt.",
+    "Needs an agent running inside a T3 thread. Create one or more ORDINARY TOP-LEVEL T3 conversations. This is not delegation and does not create child agents/subagents. For delegated work, choose models from orchestrator_capabilities. Prefer native subagents only when they support the chosen model and carry every requested or inherited model option; otherwise call delegate_task, including for same-provider work. Use create_threads for a batch of separate top-level threads sharing this checkout. Prefer t3_thread_launch for a single thread. Both require the user to request separate/new/top-level threads or conversations. Each entry may override provider, model, options, runtime mode, and interaction mode; omitted settings inherit, with model options inheriting only when the provider instance and model match. Project, branch, and worktree always inherit and cannot be overridden here. For independent implementation or a PR stack in its own worktree, use t3_thread_launch with workspaceStrategy instead of asking the agent to create a worktree in its prompt.",
   parameters: OrchestratorMcpCreateThreadsInput,
   success: OrchestratorMcpCreateThreadsResult,
   failure: OrchestratorMcpFailure,

packages/provider-core/src/server/orchestrationInstructions.test.ts

@@ -0,0 +1,24 @@
+import { describe, expect, it } from "vite-plus/test";
+
+import {
+  t3OrchestrationPromptForFirstRun,
+  t3OrchestrationSystemPrompt,
+} from "./orchestrationInstructions.ts";
+
+// Codex native spawn copies the parent's reasoning effort and service tier, so inherited
+// Codex options alone must not push plain subagent requests onto delegate_task (#23525).
+describe("orchestration instructions for inherited Codex options", () => {
+  const injected = {
+    systemPrompt: t3OrchestrationSystemPrompt(true)!,
+    firstRunPrompt: t3OrchestrationPromptForFirstRun({
+      prompt: "Summarize two packages with parallel subagents.",
+      runOrdinal: 1,
+      hasT3Mcp: true,
+    }),
+  };
+
+  it.each(Object.entries(injected))("%s lets Codex native subagents inherit options", (_, text) => {
+    expect(text).not.toMatch(/no model options are requested or inherited/i);
+    expect(text).toMatch(/Codex native subagents keep the parent's reasoning effort/i);
+  });
+});

packages/provider-core/src/server/orchestrationInstructions.ts

@@ -10,7 +10,7 @@ export const T3_CODE_ORCHESTRATION_SECTIONS: ReadonlyArray<string> = [
 
 The \`t3-code\` MCP server provides app-owned orchestration. Treat these concepts distinctly:
 
-- A delegated task/subagent is child work owned by the current thread. Use \`orchestrator_capabilities\` to discover the current provider/model IDs from the same live catalog as the composer, including configured custom models. It lists a short set of models per provider; pass \`search\` to find others. Do not treat a native tool's model list as the full list of available subagent models. Prefer native subagent tools for same-provider work only when they support the chosen model and no model options are requested or inherited. Use \`delegate_task\` with that provider instance and model when native tools cannot, including for same-provider work. Also use \`delegate_task\` for cross-provider or explicitly T3-owned child tasks. For a child task that needs a separate checkout, use \`delegate_task\` with \`isolation:"worktree"\` across any provider or model. It starts from the parent local commit and runs project setup; uncommitted edits are not copied. Retain each returned \`taskId\`, and use \`task_status\` or \`task_cancel\` to manage it. The returned \`childThreadId\` is backing storage for the subagent, not the target for starting another delegated review round.
+- A delegated task/subagent is child work owned by the current thread. Use \`orchestrator_capabilities\` to discover the current provider/model IDs from the same live catalog as the composer, including configured custom models. It lists a short set of models per provider; pass \`search\` to find others. Do not treat a native tool's model list as the full list of available subagent models. Prefer native subagent tools for same-provider work only when they support the chosen model and carry every requested or inherited model option, as described below. Use \`delegate_task\` with that provider instance and model when native tools cannot, including for same-provider work. Also use \`delegate_task\` for cross-provider or explicitly T3-owned child tasks. For a child task that needs a separate checkout, use \`delegate_task\` with \`isolation:"worktree"\` across any provider or model. It starts from the parent local commit and runs project setup; uncommitted edits are not copied. Retain each returned \`taskId\`, and use \`task_status\` or \`task_cancel\` to manage it. The returned \`childThreadId\` is backing storage for the subagent, not the target for starting another delegated review round.
 - \`t3_thread_launch\` and \`create_threads\` create ordinary top-level T3 conversations. Use them only when the user explicitly asks for separate/new/top-level threads or conversations. Never use them merely because the user said "subagent" or requested parallel delegated work.
 - For every T3 delegated review round, call \`delegate_task\` again. Include the original brief, prior findings, responses, and unresolved objections in each new task prompt. Track each round by its own \`taskId\`. Use a distinct \`clientRequestId\` per round, stable across retries of that round. Do not use \`t3_thread_send\` on \`childThreadId\` to continue a delegated review.`,
   `- \`schedule_task\` creates persistent recurring work in the app scheduler. Pass \`schedule\` as a structured object, never as JSON text: \`{"type":"interval","everyMs":3600000}\` for an interval, or \`{"type":"fixed_time","timeOfDay":"09:00","weekdays":[1,2,3,4,5]}\` for a wall-clock schedule, or \`{"type":"webhook"}\` to run on each request to the returned \`webhookUrl\` (the run sees the request only through \`{{body.path}}\`-style placeholders in the prompt). By default runs return to the current thread, which suits orchestrating: each trigger arrives here and you delegate or dedupe; set \`bindToCurrentThread=false\` only when the user wants a fresh thread for every run. After scheduling a timer, report the returned cadence and next run time; for a webhook, report its \`webhookUrl\`, or say T3 Connect remote access is needed if it is missing.
@@ -36,7 +36,7 @@ ACP fallback: some ACP agents accept the injected MCP server but fail to expose
 ### Showing visuals
 
 When a chart, table, diagram, image collage, or mockup would say more than prose, build a self-contained HTML page, check it with \`html_preview\`, then publish it with \`html_render\` before your final reply. The reader sees the page above that reply, so don't announce or restate it; add only what it doesn't say.`,
-  `- When model options are requested or the child keeps the parent's provider instance and model with nonempty \`inheritedOptions\` from \`orchestrator_capabilities\`, use \`delegate_task\` even for native-supported models. Native subagents cannot reliably apply T3 model options. Read the selected model's descriptors from \`orchestrator_capabilities\` and put settings in \`target.options\`, not only in the prompt. Claude commonly uses {"effort":"max","fastMode":true}; Codex uses {"reasoningEffort":"max","serviceTier":"priority"}. Do not translate max to high or ignore unsupported settings: report a capability mismatch. Options inherit only for the same provider instance and model; explicit entries override individual values, and [] or {} clears them. Check the returned providerInstanceId, model, and options against the requested selection.`,
+  `- Codex native subagents keep the parent's reasoning effort and supported service tier unless a role or a configured subagent default overrides them, so inherited Codex \`inheritedOptions\` alone do not require \`delegate_task\`. Use \`delegate_task\` even for native-supported models when model options are requested, when a role or subagent default would change the inherited settings, or when the child keeps the parent's provider instance and model with nonempty \`inheritedOptions\` from \`orchestrator_capabilities\` on any other provider. Other native subagents cannot reliably apply T3 model options such as Claude effort or fastMode. Read the selected model's descriptors from \`orchestrator_capabilities\` and put settings in \`target.options\`, not only in the prompt. Claude commonly uses {"effort":"max","fastMode":true}; Codex uses {"reasoningEffort":"max","serviceTier":"priority"}. Do not translate max to high or ignore unsupported settings: report a capability mismatch. Options inherit only for the same provider instance and model; explicit entries override individual values, and [] or {} clears them. Check the returned providerInstanceId, model, and options against the requested selection.`,
 ];
 
 export const T3_CODE_ORCHESTRATION_INSTRUCTIONS = `