feat(config): support object-style fallback_models with per-model settings

Add support for object-style entries in fallback_models arrays, enabling
per-model configuration of variant, reasoningEffort, temperature, top_p,
maxTokens, and thinking settings.

- Zod schema for FallbackModelObject with full validation
- normalizeFallbackModels() and flattenToFallbackModelStrings() utilities
- Provider-agnostic model resolution pipeline with fallback chain
- Session prompt params state management
- Fallback chain construction with prefix-match lookup
- Integration across delegate-task, background-agent, and plugin layers

src/tools/delegate-task/sync-prompt-sender.test.ts

@@ -3,9 +3,19 @@ const {
   test: bunTest,
   expect: bunExpect,
   mock: bunMock,
+  afterEach: bunAfterEach,
 } = require("bun:test")
 
+const {
+  clearSessionPromptParams,
+  getSessionPromptParams,
+} = require("../../shared/session-prompt-params-state")
+
 bunDescribe("sendSyncPrompt", () => {
+  bunAfterEach(() => {
+    clearSessionPromptParams("test-session")
+  })
+
   bunTest("passes question=false via tools parameter", async () => {
     //#given
     const { sendSyncPrompt } = require("./sync-prompt-sender")
@@ -214,6 +224,67 @@ bunDescribe("sendSyncPrompt", () => {
     bunExpect(promptArgs.body.variant).toBe("medium")
   })
 
+  bunTest("passes promoted fallback model settings through supported prompt channels", async () => {
+    //#given
+    const { sendSyncPrompt } = require("./sync-prompt-sender")
+
+    let promptArgs: any
+    const promptWithModelSuggestionRetry = bunMock(async (_client: any, input: any) => {
+      promptArgs = input
+    })
+
+    const input = {
+      sessionID: "test-session",
+      agentToUse: "oracle",
+      args: {
+        description: "test task",
+        prompt: "test prompt",
+        run_in_background: false,
+        load_skills: [],
+      },
+      systemContent: undefined,
+      categoryModel: {
+        providerID: "openai",
+        modelID: "gpt-5.4",
+        variant: "low",
+        reasoningEffort: "high",
+        temperature: 0.4,
+        top_p: 0.7,
+        maxTokens: 4096,
+        thinking: { type: "disabled" },
+      },
+      toastManager: null,
+      taskId: undefined,
+    }
+
+    //#when
+    await sendSyncPrompt(
+      { session: { promptAsync: bunMock(async () => ({ data: {} })) } },
+      input,
+      {
+        promptWithModelSuggestionRetry,
+        promptSyncWithModelSuggestionRetry: bunMock(async () => {}),
+      },
+    )
+
+    //#then
+    bunExpect(promptWithModelSuggestionRetry).toHaveBeenCalledTimes(1)
+    bunExpect(promptArgs.body.model).toEqual({
+      providerID: "openai",
+      modelID: "gpt-5.4",
+    })
+    bunExpect(promptArgs.body.variant).toBe("low")
+    bunExpect(promptArgs.body.options).toBeUndefined()
+    bunExpect(getSessionPromptParams("test-session")).toEqual({
+      temperature: 0.4,
+      topP: 0.7,
+      options: {
+        reasoningEffort: "high",
+        thinking: { type: "disabled" },
+        maxTokens: 4096,
+      },
+    })
+  })
   bunTest("retries with promptSync for oracle when promptAsync fails with unexpected EOF", async () => {
     //#given
     const { sendSyncPrompt } = require("./sync-prompt-sender")
@@ -289,7 +360,7 @@ bunDescribe("sendSyncPrompt", () => {
     )
 
     //#then
-    bunExpect(result).toContain("JSON Parse error: Unexpected EOF")
+    bunExpect(result).toContain("Unexpected EOF")
     bunExpect(promptWithModelSuggestionRetry).toHaveBeenCalledTimes(1)
     bunExpect(promptSyncWithModelSuggestionRetry).toHaveBeenCalledTimes(0)
   })