@ferris1225/pi-subagents 1.0.1 → 2.0.0

package/src/config.ts

@@ -17,10 +17,10 @@ import { dirname, join } from "node:path";
 import { getAgentDir, withFileMutationQueue } from "@earendil-works/pi-coding-agent";
 
 /** Full catalog of agents shipped with the package (selectable in /subagents-setup). */
-export const BUILTIN_AGENT_NAMES = ["explore", "worker", "reviewer"] as const;
+export const BUILTIN_AGENT_NAMES = ["explore", "worker", "cleaner", "reviewer"] as const;
 
-/** Agents enabled out of the box. */
-export const DEFAULT_ENABLED_AGENTS: readonly string[] = ["explore", "worker", "reviewer"];
+/** Agents enabled out of the box on a fresh install. Explicit configured lists are preserved. */
+export const DEFAULT_ENABLED_AGENTS: readonly string[] = ["explore", "worker", "cleaner", "reviewer"];
 
 export const AGENT_SCOPE_VALUES = ["user", "project", "both"] as const;
 export type AgentScope = (typeof AGENT_SCOPE_VALUES)[number];
@@ -52,24 +52,20 @@ export const MAX_FIX_ROUNDS_LIMIT = 5;
 
 /**
  * Default idle timeout in seconds: a sub-agent whose stdout (JSON event stream)
- * goes silent for this long is terminated and may be retried with the fallback
- * model. 0 disables the idle watchdog. Default: 90.
+ * goes silent for this long is terminated; a selected model then hands the
+ * retained session to current main. 0 disables the watchdog. Default: 90.
  */
 export const DEFAULT_IDLE_TIMEOUT_SEC = 90;
 /** Upper bound accepted for idleTimeoutSec (defensive clamp). 0 disables. */
 export const IDLE_TIMEOUT_SEC_LIMIT = 600;
 
 export interface SubagentsConfig {
-	/** Agent names that are discoverable and injected. Default: explore, worker, reviewer. */
+	/** Agent names that are discoverable and injected. Fresh-install default: explore, worker, cleaner, reviewer. */
 	enabledAgents: string[];
-	/** Per-agent primary model override, keyed by agent name, as "provider/model-id". */
+	/** Per-agent model override, keyed by agent name, as "provider/model-id". */
 	agentModels: Record<string, string>;
-	/** Optional per-agent backup model, tried after the primary and before the current main-window model. */
-	agentBackupModels: Record<string, string>;
-	/** Per-agent thinking-level override, keyed by agent name. */
+	/** Optional per-agent thinking preference. Runtime clamps it to the effective model's supported levels. */
 	agentThinkingLevels: Record<string, ThinkingLevel>;
-	/** Thinking level for sub-agents without a per-agent override or frontmatter default. Default: "high". */
-	thinkingLevel: ThinkingLevel;
 	/**
 	 * When a review passes (REVIEW_PASS verdict), deliver it without waking the
 	 * main agent. Disabled by default so passing reviews still resume orchestration.
@@ -98,8 +94,8 @@ export interface SubagentsConfig {
 	maxFixRounds: number;
 	/**
 	 * Idle timeout in seconds: a sub-agent whose stdout (JSON event stream) goes
-	 * silent for this long is terminated and may be retried with the fallback
-	 * model. 0 disables the idle watchdog. Default: 90.
+	 * silent for this long is terminated; a configured agent model then hands
+	 * off to the current main model. 0 disables the idle watchdog. Default: 90.
 	 */
 	idleTimeoutSec: number;
 	/**
@@ -119,9 +115,7 @@ export interface SubagentsConfig {
 export const DEFAULT_CONFIG: SubagentsConfig = {
 	enabledAgents: [...DEFAULT_ENABLED_AGENTS],
 	agentModels: {},
-	agentBackupModels: {},
 	agentThinkingLevels: {},
-	thinkingLevel: DEFAULT_THINKING_LEVEL,
 	notifyOnReviewPass: false,
 	maxResultLines: DEFAULT_MAX_RESULT_LINES,
 	proactiveInjection: true,
@@ -179,12 +173,6 @@ export function normalizeConfig(raw: unknown): SubagentsConfig {
 		}
 	}
 
-	if (isRecord(raw.agentBackupModels)) {
-		for (const [key, value] of Object.entries(raw.agentBackupModels)) {
-			if (isModelReference(value)) config.agentBackupModels[key.trim()] = value.trim();
-		}
-	}
-
 	if (isRecord(raw.agentThinkingLevels)) {
 		for (const [key, value] of Object.entries(raw.agentThinkingLevels)) {
 			if (
@@ -196,10 +184,6 @@ export function normalizeConfig(raw: unknown): SubagentsConfig {
 		}
 	}
 
-	if (typeof raw.thinkingLevel === "string" && (THINKING_LEVEL_VALUES as readonly string[]).includes(raw.thinkingLevel)) {
-		config.thinkingLevel = raw.thinkingLevel as ThinkingLevel;
-	}
-
 	if (typeof raw.notifyOnReviewPass === "boolean") {
 		config.notifyOnReviewPass = raw.notifyOnReviewPass;
 	}
@@ -246,7 +230,6 @@ function defaultConfig(): SubagentsConfig {
 		...DEFAULT_CONFIG,
 		enabledAgents: [...DEFAULT_CONFIG.enabledAgents],
 		agentModels: {},
-		agentBackupModels: {},
 		agentThinkingLevels: {},
 		announcedFeatures: [],
 	};

package/src/dispatch.ts

@@ -1,12 +1,11 @@
 /**
- * The `subagent` tool: dispatches explore/worker/reviewer agents as isolated pi
+ * The `subagent` tool: dispatches explore/worker/cleaner/reviewer agents as isolated pi
  * child processes, single or parallel. Owns the dispatch pipeline: config load,
- * per-agent model-pool resolution, per-run status tracking, the auto-fix chain
+ * per-agent selected→main model routing, per-run status tracking, the auto-fix chain
  * (REVIEW_FAIL → worker → re-review), and completion delivery.
  *
- * Vision: a task flagged `vision: true` uses the configured vision model as an
- * explicit primary, then the agent's configured backup and the current
- * main-window model. Stale refs remain in the pool and fail normally at runtime.
+ * Vision: a task flagged `vision: true` uses the configured vision model, then
+ * hands directly to the current main-window model on model/provider failure.
  */
 
 import { StringEnum } from "@earendil-works/pi-ai";
@@ -21,7 +20,12 @@ import {
 	completionTriggersTurn,
 	type CompletionMessageItem,
 } from "./completion.ts";
-import { loadConfig, type SubagentsConfig } from "./config.ts";
+import {
+	DEFAULT_THINKING_LEVEL,
+	loadConfig,
+	type SubagentsConfig,
+	type ThinkingLevel,
+} from "./config.ts";
 import {
 	dispatchFailedResult,
 	failedStartResult,
@@ -37,7 +41,14 @@ import {
 	shouldTriggerFixLoop,
 	type ChainStep,
 } from "./fixloop.ts";
-import { currentModelRef, resolveAgentModelPool } from "./models.ts";
+import {
+	availableModelsInScope,
+	currentModelRef,
+	findModelByRef,
+	modelRef,
+	resolveAgentModelRoute,
+	resolveThinkingLevel,
+} from "./models.ts";
 import {
 	formatTaskSummary,
 	formatToolActivity,
@@ -49,14 +60,13 @@ import type { SubagentRuntime, SubagentThread, ThreadState } from "./runtime.ts"
 import { persistRecoveryRecords, recoveryRecordFromFinalization } from "./recovery.ts";
 import { forkRetainedSession } from "./session-fork.ts";
 import {
-	buildFallbackResumeReason,
 	buildResumePrompt,
 	RpcRunControl,
 	getResultOutput,
 	isFailedResult,
 	isModelLevelFailure,
 	reviewVerdict,
-	runSingleAgentWithModelFallback,
+	runSingleAgentWithMainFallback,
 	type SingleResult,
 	type SubagentDetails,
 	type SubagentLiveEvent,
@@ -86,14 +96,13 @@ interface DispatchEnvironment {
 	ctx: ExtensionContext;
 	config: SubagentsConfig;
 	agents: AgentConfig[];
-	sessionRef?: string;
 }
 
 const VISION_DESCRIPTION =
-	"Set true when the task may require viewing images (screenshots, mockups, designs) — the configured vision model becomes primary, followed by the agent backup and current main-window model";
+	"Set true when the task may require viewing images (screenshots, mockups, designs) — the configured vision model is used first, then model-level failures hand directly to the current main-window model";
 
 const ISOLATION_DESCRIPTION =
-	"Filesystem isolation: shared uses the caller's working tree; worktree creates a detached temporary Git worktree (write-capable agents only)";
+	"Filesystem isolation: shared uses the caller's working tree; worktree creates a detached temporary Git worktree (write-capable agents, including worker and cleaner, only)";
 
 const IsolationSchema = Type.Optional(
 	StringEnum(["shared", "worktree"] as const, { description: ISOLATION_DESCRIPTION }),
@@ -175,21 +184,39 @@ function serializeAutoFixChain(
 	};
 }
 
-function resolveDispatchModelPool(
+interface DispatchModelRoute {
+	agent: AgentConfig;
+	mainFallbackRef?: string;
+	thinkingLevel: ThinkingLevel;
+	thinkingLevelForModel: (ref?: string) => ThinkingLevel;
+}
+
+function resolveDispatchModelRoute(
 	agent: AgentConfig,
 	config: SubagentsConfig,
-	mainRef: string | undefined,
+	ctx: ExtensionContext,
 	vision: boolean,
-): { agent: AgentConfig; fallbackModelRefs: string[] } {
-	const pool = resolveAgentModelPool({
-		primaryRef: vision ? config.visionModel : config.agentModels[agent.name],
-		backupRef: config.agentBackupModels[agent.name],
+): DispatchModelRoute {
+	const availableModels = availableModelsInScope(ctx);
+	const mainRef = currentModelRef(ctx);
+	const route = resolveAgentModelRoute({
+		selectedRef: vision ? config.visionModel : config.agentModels[agent.name],
 		mainRef,
 		declaredDefaultRef: agent.model,
+		availableRefs: availableModels.map(modelRef),
 	});
+	const preferred = config.agentThinkingLevels[agent.name] ?? agent.thinking ?? DEFAULT_THINKING_LEVEL;
+	const thinkingLevelForModel = (ref?: string): ThinkingLevel => {
+		const model = ref === mainRef && ctx.model
+			? ctx.model
+			: findModelByRef(availableModels, ref);
+		return resolveThinkingLevel(model, preferred);
+	};
 	return {
-		agent: { ...agent, model: pool.primaryRef },
-		fallbackModelRefs: pool.fallbackModelRefs,
+		agent: { ...agent, model: route.primaryRef },
+		mainFallbackRef: route.mainFallbackRef,
+		thinkingLevel: thinkingLevelForModel(route.primaryRef),
+		thinkingLevelForModel,
 	};
 }
 
@@ -199,28 +226,31 @@ export function registerSubagentTool(pi: ExtensionAPI, runtime: SubagentRuntime)
 		label: "Subagent",
 		description: [
 			"Delegate a discrete, self-contained task to a specialized sub-agent running in an ISOLATED context window.",
-			"Agents: explore (read-only codebase recon), worker (implement/fix/refactor/test, full tools), reviewer (adversarial pre-commit review, read-only).",
+			"Built-in roles (only configured enabled agents can dispatch): explore (read-only codebase recon), worker (implement/fix/refactor/test, full tools), cleaner (explicit evidence-first cleanup, full/write tools), reviewer (adversarial pre-commit gate, read-only).",
+			"When cleaner is enabled, route explicit cleanup intent in any language (for example dead code, redundancy, simplification, or over-engineering), including a requested periodic cleanup pass. Audit/find/inspect/report is read-only evidence, while explicit remove/clean/simplify/refactor permits verified edits. Generic code review goes to reviewer. Never route cleaner by PR count or as the pre-commit gate; reviewer reviews its edits.",
 			"Modes: single ({agent, task}) or parallel ({tasks: [{agent, task}, ...]}).",
-			"Isolation: single tasks default to shared; parallel worker tasks default to detached Git worktrees unless isolation: shared is explicit. explore/reviewer cannot use worktree isolation.",
+			"Isolation: single tasks default to shared; parallel worker tasks default to detached Git worktrees unless isolation: shared is explicit. Cleaner is write-capable and can opt into worktree isolation; explore/reviewer cannot use it.",
 			"Use subagent_control to steer, retarget, park, resume, or fork a thread by its stable run id.",
 			"It starts agents in the background and immediately returns control to the main window; completion messages automatically wake the main agent to continue.",
 			"Each agent has no memory of this conversation — brief it fully (goal, exact paths, constraints, expected output).",
 			"Results arrive as wake-up messages automatically — you do NOT need to wait. If you must get a result in-turn, subagent_wait is a non-blocking lookup by default (pass timeoutMs to block).",
-			"Vision: set vision: true when the task may require viewing images (screenshots, mockups, design files — e.g. frontend work) — the configured vision model is primary, followed by that agent's backup and the current main-window model.",
+			"Vision: set vision: true when the task may require viewing images (screenshots, mockups, design files — e.g. frontend work) — the configured vision model is used first, then model-level failures hand directly to the current main-window model.",
 		].join(" "),
 		promptSnippet:
-			"Start background subagents: explore (read-only search), worker (implement), reviewer (adversarial review); completion automatically resumes the main agent. Simple tasks: use direct tools, not subagents.",
+			"Start background subagents: explore (read-only search), worker (implement), cleaner (explicit evidence-first cleanup), reviewer (pre-commit review); completion automatically resumes the main agent. Simple tasks: use direct tools, not subagents.",
 		promptGuidelines: [
-			"Delegate only when an isolated context genuinely pays: broad exploration, a self-contained implementation, or a review gate. Handle simple lookups and one-line edits inline with direct tools — never spawn a sub-agent for them.",
+			"Delegate only when an isolated context genuinely pays: broad exploration, a self-contained implementation, explicit evidence-first cleanup, or a review gate. Handle simple lookups and one-line edits inline with direct tools — never spawn a sub-agent for them.",
 			"Use subagent with agent 'explore' for broad or open-ended code search before large changes; a targeted 'where is X' is a direct grep/read.",
+			"Treat explore output as a retrieval index, not authority: re-read load-bearing files before editing or deciding deletion, security, compatibility, persistence, or dynamic reachability. The cheapest model can cost more through rework on complex dynamic, concurrent, migration, or security-sensitive code; choose a stronger model or specialist there.",
 			"Use subagent with agent 'worker' for a self-contained implementation task worth a separate context; it plans internally.",
-			"Use subagent with agent 'reviewer' for a fresh read-only review before reporting work done or committing.",
+			"When cleaner is enabled, use subagent with agent 'cleaner' only for explicit cleanup intent in any language (for example dead code, redundancy, simplification, or over-engineering) or a requested periodic cleanup pass. Audit/find/inspect/report means read-only ranked evidence; apply only for explicit remove/clean/simplify/refactor wording. Generic code review goes to reviewer. Never trigger cleaner from PR count or as a pre-commit gate; send non-trivial cleaner edits to reviewer.",
+			"Use subagent with agent 'reviewer' for the fresh read-only gate before reporting non-trivial work done or committing, including after cleaner edits.",
 			"subagent launches work in the background and ends the current turn; when a result arrives, the main agent is automatically resumed with it.",
 			"Run independent tasks in parallel by passing a tasks array to subagent; parallel worker items default to isolation: worktree so their edits are integrated independently. Pass isolation: shared only when workers intentionally need the caller's live uncommitted tree.",
-			"Use isolation: worktree only for worker/write-capable agents and only inside a Git repository with a committed HEAD; setup or integration failures never silently fall back to shared.",
+			"Use isolation: worktree only for worker, cleaner, or another write-capable agent and only inside a Git repository with a committed HEAD; parallel worker tasks default to worktree, while cleaner must opt in. Setup or integration failures never silently fall back to shared.",
 			"NEVER sleep or poll, and do NOT call subagent_wait to hold the turn — subagent ends the turn immediately and the result arrives as a message that wakes you automatically (even mid-turn). Ending your turn is the default and the only correct way to wait.",
 			"If you must keep the turn for a result, call subagent_wait with an explicit timeoutMs (non-blocking by default) — never bash sleep/timeout to wait for a sub-agent.",
-			"When a delegated task may require viewing images (frontend screenshots, mockups, design comparisons), pass vision: true and give the sub-agent the exact image paths — it reads them with its read tool. The configured vision model becomes primary; model-level failures continue through the agent's backup pool and current main-window model.",
+			"When a delegated task may require viewing images (frontend screenshots, mockups, design comparisons), pass vision: true and give the sub-agent the exact image paths — it reads them with its read tool. The configured vision model is used first; model-level failures hand directly to the current main-window model.",
 			"When a sub-agent result arrives it is already shown to the user — do NOT restate, paraphrase, or summarize it; reply with only your own conclusion or next action (often just one line), since duplicating the result wastes tokens for nothing.",
 		],
 		parameters: SubagentParams,
@@ -267,6 +297,7 @@ export function registerSubagentTool(pi: ExtensionAPI, runtime: SubagentRuntime)
 							break;
 						case "model":
 							monitor.setModel(runId, e.model, e.fallbackFrom);
+							monitor.setThinking(runId, e.thinking);
 							break;
 						case "usage":
 							monitor.setUsage(runId, e.usage, e.model);
@@ -292,7 +323,6 @@ export function registerSubagentTool(pi: ExtensionAPI, runtime: SubagentRuntime)
 				enabledNames: config.enabledAgents,
 				projectTrusted: ctx.isProjectTrusted?.() === true,
 			});
-			const sessionRef = currentModelRef(ctx);
 			const agents = discovery.agents;
 
 			const hasTasks = (params.tasks?.length ?? 0) > 0;
@@ -356,32 +386,35 @@ export function registerSubagentTool(pi: ExtensionAPI, runtime: SubagentRuntime)
 			): Promise<{ runId?: number; result: SingleResult }> => {
 				const agent = agents.find((candidate) => candidate.name === agentName);
 				if (!agent) return { result: failedStartResult(agentName, task, `Unknown agent: "${agentName}".`) };
-				// Vision chains keep the vision override as each round's primary while
-				// retaining that worker/reviewer's own configured backup pool.
-				const pool = resolveDispatchModelPool(agent, config, sessionRef, vision);
-				const thinkingLevel = config.agentThinkingLevels[agent.name] ?? agent.thinking ?? config.thinkingLevel;
-				const runId = monitor.addRun(agent.name, task, pool.agent.model, thinkingLevel, meta);
+				// Vision chains use the vision override first; every model-level failure
+				// hands directly to the current main model with re-clamped thinking.
+				const route = resolveDispatchModelRoute(agent, config, ctx, vision);
+				const thinkingLevel = route.thinkingLevel;
+				const runId = monitor.addRun(agent.name, task, route.agent.model, thinkingLevel, meta);
 				const onLive = makeLiveHandler(runId);
 				try {
-					const result = await runSingleAgentWithModelFallback(
+					const result = await runSingleAgentWithMainFallback(
 						{
 							defaultCwd: executionCwd,
 							cwd: executionCwd,
-							agent: pool.agent,
+							agent: route.agent,
 							agentName,
 							task,
 							thinkingLevel,
+							thinkingLevelForModel: route.thinkingLevelForModel,
 							signal,
 							onLive,
 							makeDetails: makeDetails("single", true),
 							idleTimeoutMs: config.idleTimeoutSec * 1000,
 						},
-						pool.fallbackModelRefs,
+						route.mainFallbackRef,
 					);
 					result.runId = runId;
+					result.projectCwd = executionCwd;
 					result.isolation = "shared";
 					runtime.retainSession(result);
 					monitor.setModel(runId, result.model, result.modelFallbackFrom);
+					monitor.setThinking(runId, result.thinking);
 					// The parent row represents the chain. Internal rounds leave live
 					// status as soon as they settle; their reports remain addressable by id.
 					finishRun(runId, isFailedResult(result) ? "failed" : "done", { silent: true });
@@ -391,8 +424,9 @@ export function registerSubagentTool(pi: ExtensionAPI, runtime: SubagentRuntime)
 					finishRun(runId, "failed", { silent: true });
 					const errorMessage = error instanceof Error ? error.message : String(error);
 					const crashed: SingleResult = {
-						...queuedResult(pool.agent, task, thinkingLevel),
+						...queuedResult(route.agent, task, thinkingLevel),
 						runId,
+						projectCwd: executionCwd,
 						isolation: "shared",
 						exitCode: 1,
 						stderr: errorMessage,
@@ -646,8 +680,6 @@ export function registerSubagentTool(pi: ExtensionAPI, runtime: SubagentRuntime)
 				worktree?: WorktreeIsolation;
 				forkedFromRunId?: number;
 				forkObjective?: string;
-				modelPool?: string[];
-				thinkingLevel?: SubagentThread["thinkingLevel"];
 			}
 
 			interface ResumeReservation {
@@ -703,12 +735,11 @@ export function registerSubagentTool(pi: ExtensionAPI, runtime: SubagentRuntime)
 				const runCtx = environment?.ctx ?? ctx;
 				const runConfig = environment?.config ?? config;
 				const runAgents = environment?.agents ?? agents;
-				const runSessionRef = environment?.sessionRef ?? sessionRef;
 				const agent = runAgents.find((candidate) => candidate.name === agentName);
 				if (!agent) return failedStartResult(agentName, task, `Unknown agent: "${agentName}".`);
 				if (isolation === "worktree" && !isWorktreeCapableAgent(agent)) {
 					return {
-						...failedStartResult(agentName, task, `Agent "${agentName}" is read-only; worktree isolation is available only to worker/write-capable agents.`),
+						...failedStartResult(agentName, task, `Agent "${agentName}" is read-only; worktree isolation is available only to write-capable agents such as worker or cleaner.`),
 						isolation,
 					};
 				}
@@ -736,36 +767,29 @@ export function registerSubagentTool(pi: ExtensionAPI, runtime: SubagentRuntime)
 					}
 				}
 				const executionCwd = worktree?.cwd ?? originalCwd;
-				const resolvedPool = resolveDispatchModelPool(agent, runConfig, runSessionRef, vision);
-				const inheritedPool = seed?.modelPool?.filter((ref) => ref.trim().length > 0) ?? [];
-				const rawPool = inheritedPool.length > 0
-					? {
-						agent: { ...agent, model: inheritedPool[0] },
-						fallbackModelRefs: inheritedPool.slice(1),
-					}
-					: resolvedPool;
+				const resolvedRoute = resolveDispatchModelRoute(agent, runConfig, runCtx, vision);
 				// Isolation is a persistent system-level invariant, not a one-shot task
-				// prefix: queued retargets, live retargets, resumes, and model fallbacks
-				// all keep the same worktree boundary.
-				const pool = isolation === "worktree"
-					? { ...rawPool, agent: withWorktreeSystemPrompt(rawPool.agent) }
-					: rawPool;
-				const thinkingLevel = seed?.thinkingLevel ?? runConfig.agentThinkingLevels[agent.name] ?? agent.thinking ?? runConfig.thinkingLevel;
-				const modelPool = [pool.agent.model, ...pool.fallbackModelRefs].filter((ref): ref is string => Boolean(ref));
+				// prefix: queued retargets, live retargets, resumes, and main-model
+				// handoffs all keep the same worktree boundary.
+				const route = isolation === "worktree"
+					? { ...resolvedRoute, agent: withWorktreeSystemPrompt(resolvedRoute.agent) }
+					: resolvedRoute;
+				const thinkingLevel = route.thinkingLevel;
 				const priorTask = existingThread?.task;
 				const priorSessionId = seed?.sessionId ?? existingThread?.sessionId;
 				const priorSessionDir = seed?.sessionDir ?? existingThread?.sessionDir;
 				if (existingThread && resumeReservation && !ownsResumeReservation(existingThread, resumeReservation)) {
 					return failedStartResult(agentName, task, `Run #${existingThread.id} changed while resume was preparing; no new generation was started.`);
 				}
-				const runId = existingThread?.id ?? monitor.addRun(agent.name, task, pool.agent.model, thinkingLevel, {
+				const runId = existingThread?.id ?? monitor.addRun(agent.name, task, route.agent.model, thinkingLevel, {
 					isolation,
 					...(seed?.forkedFromRunId !== undefined ? { forkedFromRunId: seed.forkedFromRunId } : {}),
 				});
 				const generation = (existingThread?.generation ?? 0) + 1;
 				const pending: SingleResult = {
-					...queuedResult(pool.agent, task, thinkingLevel),
+					...queuedResult(route.agent, task, thinkingLevel),
 					runId,
+					projectCwd: originalCwd,
 					isolation,
 					...(isolation === "worktree" ? { integrationStatus: "pending" as const } : {}),
 					...(seed?.sessionId && seed.sessionDir
@@ -774,7 +798,7 @@ export function registerSubagentTool(pi: ExtensionAPI, runtime: SubagentRuntime)
 					...(seed?.forkedFromRunId !== undefined ? { forkedFromRunId: seed.forkedFromRunId } : {}),
 				};
 				if (existingThread) {
-					monitor.restartRun(runId, agent.name, task, pool.agent.model, thinkingLevel, isolation);
+					monitor.restartRun(runId, agent.name, task, route.agent.model, thinkingLevel, isolation);
 					runtime.settledRuns.delete(runId);
 				}
 
@@ -810,7 +834,6 @@ export function registerSubagentTool(pi: ExtensionAPI, runtime: SubagentRuntime)
 					thread.cwd = originalCwd;
 					thread.executionCwd = executionCwd;
 					thread.vision = vision;
-					thread.modelPool = modelPool;
 					thread.thinkingLevel = thinkingLevel;
 					thread.isolation = isolation;
 					thread.worktree = worktree;
@@ -835,7 +858,6 @@ export function registerSubagentTool(pi: ExtensionAPI, runtime: SubagentRuntime)
 						cwd: originalCwd,
 						executionCwd,
 						vision,
-						modelPool,
 						thinkingLevel,
 						isolation,
 						worktree,
@@ -859,7 +881,7 @@ export function registerSubagentTool(pi: ExtensionAPI, runtime: SubagentRuntime)
 				thread.notifyIsolationFailure = (finalization) => {
 					const paths = [finalization.worktreePath, finalization.patchPath].filter(Boolean).join(" · ");
 					runCtx.ui.notify(
-						`✗ worker worktree ${finalization.integrated ? "cleanup" : "integration"} failed${paths ? ` · retained ${paths}` : ""}: ${finalization.error ?? "unknown Git integration error"}`,
+						`✗ ${agent.name} worktree ${finalization.integrated ? "cleanup" : "integration"} failed${paths ? ` · retained ${paths}` : ""}: ${finalization.error ?? "unknown Git integration error"}`,
 						"error",
 					);
 				};
@@ -1122,7 +1144,6 @@ export function registerSubagentTool(pi: ExtensionAPI, runtime: SubagentRuntime)
 								ctx: currentCtx,
 								config: currentConfig,
 								agents: currentAgents,
-								sessionRef: currentModelRef(currentCtx),
 							},
 							seed,
 							reservation,
@@ -1285,7 +1306,6 @@ export function registerSubagentTool(pi: ExtensionAPI, runtime: SubagentRuntime)
 								ctx: currentCtx,
 								config: currentConfig,
 								agents: currentAgents,
-								sessionRef: currentModelRef(currentCtx),
 							},
 							{
 								sessionId: forkedSession.sessionId,
@@ -1294,8 +1314,6 @@ export function registerSubagentTool(pi: ExtensionAPI, runtime: SubagentRuntime)
 								worktree: childWorktree,
 								forkedFromRunId: runId,
 								forkObjective,
-								modelPool: [...thread.modelPool],
-								thinkingLevel: thread.thinkingLevel,
 							},
 						);
 						if (child.exitCode !== -1 || child.runId === undefined) {
@@ -1345,14 +1363,15 @@ export function registerSubagentTool(pi: ExtensionAPI, runtime: SubagentRuntime)
 						if (runtime.threads.get(runId)?.generation !== generation) return;
 						let result: SingleResult;
 						try {
-							result = await runSingleAgentWithModelFallback(
+							result = await runSingleAgentWithMainFallback(
 								{
 									defaultCwd: executionCwd,
-									agent: pool.agent,
+									agent: route.agent,
 									agentName,
 									task,
 									cwd: executionCwd,
 									thinkingLevel,
+									thinkingLevelForModel: route.thinkingLevelForModel,
 									signal: backgroundSignal,
 									onLive,
 									control,
@@ -1364,11 +1383,11 @@ export function registerSubagentTool(pi: ExtensionAPI, runtime: SubagentRuntime)
 											sessionDir: priorSessionDir,
 											stdinText: seed?.prompt ?? (newObjectiveOnResume
 												? task
-												: buildResumePrompt(priorTask ?? task, buildFallbackResumeReason())),
+												: buildResumePrompt(priorTask ?? task, "the retained thread was resumed")),
 										}
 										: {}),
 								},
-								pool.fallbackModelRefs,
+								route.mainFallbackRef,
 							);
 						} catch (error) {
 							const errorMessage = error instanceof Error ? error.message : String(error);
@@ -1387,6 +1406,7 @@ export function registerSubagentTool(pi: ExtensionAPI, runtime: SubagentRuntime)
 						// no monitor mutation, result registration, or completion delivery.
 						if (runtime.threads.get(runId)?.generation !== generation) return;
 						result.runId = runId;
+						result.projectCwd = originalCwd;
 						result.isolation = isolation;
 						result.forkedFromRunId = thread.forkedFromRunId;
 						result.forkChildRunIds = [...thread.forkChildRunIds];
@@ -1398,6 +1418,7 @@ export function registerSubagentTool(pi: ExtensionAPI, runtime: SubagentRuntime)
 						thread.lastResult = result;
 						runtime.retainSession(result);
 						monitor.setModel(runId, result.model, result.modelFallbackFrom);
+						monitor.setThinking(runId, result.thinking);
 
 						// Destructive stop owns publication once it has synchronously claimed
 						// the lifecycle. Leave the partial result/session on the thread; the
@@ -1457,8 +1478,8 @@ export function registerSubagentTool(pi: ExtensionAPI, runtime: SubagentRuntime)
 							const completion: CompletionMessageItem = {
 								agent: result.agent,
 								block: modelLevel
-									? `${formatCompletionBlock(result, runConfig.maxResultLines, runCtx.cwd)}\n\n${modelLevelTakeoverNote(result, { runId })}`
-									: formatCompletionBlock(result, runConfig.maxResultLines, runCtx.cwd),
+									? `${formatCompletionBlock(result, runConfig.maxResultLines, result.projectCwd ?? originalCwd)}\n\n${modelLevelTakeoverNote(result, { runId })}`
+									: formatCompletionBlock(result, runConfig.maxResultLines, result.projectCwd ?? originalCwd),
 								triggerTurn: completionTriggersTurn(result, runConfig.notifyOnReviewPass),
 							};
 							if (modelLevel) {
@@ -1512,7 +1533,7 @@ export function registerSubagentTool(pi: ExtensionAPI, runtime: SubagentRuntime)
 							!thread.retired;
 						try {
 							const crashed: SingleResult = {
-								...dispatchFailedResult(pool.agent, control.getObjective(), error, thinkingLevel),
+								...dispatchFailedResult(route.agent, control.getObjective(), error, thinkingLevel),
 								runId,
 								isolation,
 								forkedFromRunId: thread.forkedFromRunId,
@@ -1531,7 +1552,7 @@ export function registerSubagentTool(pi: ExtensionAPI, runtime: SubagentRuntime)
 								runtime.sendCompletionGroup([
 									{
 										agent: agent.name,
-										block: formatCompletionBlock(crashed, runConfig.maxResultLines, runCtx.cwd),
+										block: formatCompletionBlock(crashed, runConfig.maxResultLines, crashed.projectCwd ?? originalCwd),
 										triggerTurn: true,
 									},
 								]);
@@ -1656,7 +1677,7 @@ export function registerSubagentTool(pi: ExtensionAPI, runtime: SubagentRuntime)
 				const pending = r.exitCode === -1;
 				const icon = statusIcon(pending ? "running" : isFailedResult(r) ? "failed" : "done", theme);
 				const usage = formatUsage(r.usage);
-				const model = `${r.model ?? "?"}${r.modelFallbackFrom ? ` (pool fallback from ${r.modelFallbackFrom})` : ""}`;
+				const model = `${r.model ?? "?"}${r.modelFallbackFrom ? ` (main after ${r.modelFallbackFrom} failed)` : ""}`;
 				const isolation = r.isolation === "worktree" ? ` · worktree ${r.integrationStatus ?? "active"}` : "";
 				const runId = r.runId === undefined ? "" : `${theme.fg("dim", `#${r.runId}`)} `;
 				const line = `${theme.fg("toolTitle", theme.bold("subagent "))}${icon} ${runId}${theme.fg("accent", r.agent)} ${theme.fg("dim", `· ${model}${r.thinking ? ` · thinking ${r.thinking}` : ""}${isolation}${pending ? " · background" : ""}${usage ? ` · ${usage}` : ""}`)}`;
@@ -1671,7 +1692,7 @@ export function registerSubagentTool(pi: ExtensionAPI, runtime: SubagentRuntime)
 				const pending = r.exitCode === -1;
 				const icon = statusIcon(pending ? "running" : isFailedResult(r) ? "failed" : "done", theme);
 				const usage = formatUsage(r.usage);
-				const model = `${r.model ?? "?"}${r.modelFallbackFrom ? ` (pool fallback from ${r.modelFallbackFrom})` : ""}`;
+				const model = `${r.model ?? "?"}${r.modelFallbackFrom ? ` (main after ${r.modelFallbackFrom} failed)` : ""}`;
 				const isolation = r.isolation === "worktree" ? ` · worktree ${r.integrationStatus ?? "active"}` : "";
 				const runId = r.runId === undefined ? "" : `${theme.fg("dim", `#${r.runId}`)} `;
 				lines.push(`  ${icon} ${runId}${theme.fg("accent", r.agent)} ${theme.fg("dim", `· ${model}${r.thinking ? ` · thinking ${r.thinking}` : ""}${isolation}${pending ? " · background" : ""}${usage ? ` · ${usage}` : ""}`)}`);