pi-subagents 0.60.0 → 0.61.0

package/src/runs/background/async-execution.ts

@@ -26,7 +26,7 @@ import { backgroundProcessOptions } from "../shared/background-process-options.t
 import { buildSkillInjection, normalizeSkillInput, resolveSkillsWithFallback } from "../../agents/skills.ts";
 import { buildAgentMemoryInjection } from "../../agents/agent-memory.ts";
 import { PI_CODING_AGENT_PACKAGE_ROOT_ENV, PROMPT_REDACTED, resolveChildCwd } from "../../shared/utils.ts";
-import { buildModelCandidates, inheritsParentModel, resolveEffectiveSubagentModel, resolveSubagentModelOverride, type AvailableModelInfo, type ParentModel } from "../shared/model-fallback.ts";
+import { buildModelCandidates, resolveEffectiveSubagentModel, resolveModelOrigin, resolveSubagentModelOverride, type AvailableModelInfo, type ModelOrigin, type ParentModel } from "../shared/model-fallback.ts";
 import { resolveToolTimeoutMs, toolTimeoutFromEnv } from "../shared/tool-timeout.ts";
 import { resolveModelScopesForAgent, type ModelScopeConfig } from "../shared/model-scope.ts";
 import { findModelInfo, resolveEffectiveThinking } from "../../shared/model-info.ts";
@@ -235,6 +235,7 @@ interface AsyncSingleParams {
 	structuredOutputSchema?: JsonSchemaObject;
 	modelOverride?: string;
 	modelOverrideFromParent?: boolean;
+	modelOrigin?: ModelOrigin;
 	fast?: boolean;
 	thinkingOverride?: AgentConfig["thinking"];
 	availableModels?: AvailableModelInfo[];
@@ -339,15 +340,15 @@ export function formatAsyncStartedMessage(headline: string, interactive: boolean
 	const guidance = interactive
 		? [
 			"The async run is detached and running in the background.",
-			"You are in an interactive session. By default, return control to the user now; Pi will wake you on completion when the run finishes or needs attention. Do NOT call subagent_wait() merely to wait, and do not run sleep/polling loops to wait for it.",
-			"When you need an explicit wake for one known run but do not need same-turn results, call subagent_wait({ id: \"...\", nonBlocking: true }) to arm a subscription and return immediately.",
-			"Override the default and call blocking subagent_wait() before ending the turn only when the current request is run-to-completion — for example, the user asked you to report results back here before continuing, or a skill must finish in one turn. In that case, call subagent_wait() to block until the run completes so its results are delivered in this turn instead of deferred.",
+			"You are in an interactive session. Return control to the user now; Pi will wake you through the native completion notification when this subagent completes or needs attention. Do not run sleep/polling loops to wait for this async subagent; it does not need a wait call.",
+			"Use bg_wait only for provider, detached, or other background work that lacks a native completion notification.",
+			"If the current turn must receive results from work without a native notification before it ends, call blocking bg_wait(); ordinary async subagent runs do not need a wait call because their completion is delivered natively.",
 			"Otherwise, continue any independent work or return control to the user. Use subagent({ action: \"status\", id: \"...\" }) for a one-shot status/result or to inspect a blocked/stale run, never as a wait loop.",
 		]
 		: [
 			"The async run is detached. Do not run sleep timers or polling loops just to wait for it.",
-			"This is a non-interactive run: Pi auto-drains current-session background work at agent_end so detached children are not abandoned; call subagent_wait() when this turn must receive the run's results before it ends, otherwise let the headless auto-drain finish the work.",
-			"Use subagent({ action: \"status\", id: \"...\" }) when you need a one-shot status/result or to inspect a blocked/stale run. To block until completion, use subagent_wait() — do not poll in a loop.",
+			"This is a non-interactive run: Pi auto-drains current-session subagent work at agent_end so detached children are not abandoned. Use bg_wait only when this turn must receive provider, detached, or other background-work results that have no native completion notification.",
+			"Use subagent({ action: \"status\", id: \"...\" }) when you need a one-shot status/result or to inspect a blocked/stale run; do not poll in a loop.",
 		];
 	return [headline, "", ...guidance].join("\n");
 }
@@ -857,14 +858,15 @@ export function buildAsyncRunnerSteps(id: string, params: AsyncRunnerStepBuildPa
 		const task = namespaceOutputPath ? taskText : injectSingleOutputInstruction(taskText, outputPath, a);
 
 		const modelScopes = resolveModelScopesForAgent(ctx.modelScope, a.name, ctx.currentModel);
-		const primaryModelFromParent = inheritsParentModel(s.model, a.model, ctx.currentModel);
+		const modelOrigin = resolveModelOrigin({ explicitModel: s.model, agentModel: a.model, parentModel: ctx.currentModel });
+		const primaryModelFromParent = modelOrigin === "inherited";
 		const primaryModel = externalRunner ? undefined : resolveEffectiveSubagentModel(
 			s.model,
 			a.model,
 			ctx.currentModel,
 			availableModels,
 			a.modelProvider ?? ctx.currentModelProvider,
-			{ scope: modelScopes },
+			{ scope: modelScopes, source: modelOrigin === "explicit" ? "explicit" : "inherited" },
 		);
 		const thinkingOverride = flatIndex === undefined ? undefined : thinkingOverridesByFlatIndex?.[flatIndex];
 		const effectiveThinking = externalRunner ? undefined : thinkingOverride ?? a.thinking;
@@ -884,15 +886,17 @@ export function buildAsyncRunnerSteps(id: string, params: AsyncRunnerStepBuildPa
 		}
 		const agentContract = s.agentContract ?? params.agentContract;
 		const permissionRules = resolvePermissionRules(ctx.permissions, a.permissions);
-		const modelCandidates = externalRunner ? [] : buildModelCandidates(primaryModel, a.fallbackModels, availableModels, a.modelProvider ?? ctx.currentModelProvider, {
-			scope: modelScopes,
-			primaryModelFromParent,
-		}).flatMap((candidate) => {
-			const resolved = applyThinkingSuffix(candidate, effectiveThinking, thinkingOverride !== undefined);
-			return resolved ? [resolved] : [];
-		});
+		let modelCandidates: string[] = [];
 		if (!externalRunner) {
 			try {
+				modelCandidates = buildModelCandidates(primaryModel, a.fallbackModels, availableModels, a.modelProvider ?? ctx.currentModelProvider, {
+					scope: modelScopes,
+					primaryModelFromParent,
+					origin: modelOrigin,
+				}).flatMap((candidate) => {
+					const resolved = applyThinkingSuffix(candidate, effectiveThinking, thinkingOverride !== undefined);
+					return resolved ? [resolved] : [];
+				});
 				for (const candidate of modelCandidates) assertThinkingWithinCeiling({ model: candidate, configThinking: effectiveThinking, ceiling: thinkingCeiling, agent: a.name, runId: id });
 			} catch (error) {
 				throw new AsyncStartValidationError(error instanceof Error ? error.message : String(error));
@@ -1588,15 +1592,27 @@ export function executeAsyncSingle(
 		: "";
 	const taskText = readsInstruction + taskWithOutputInstruction;
 	const modelScopes = resolveModelScopesForAgent(ctx.modelScope, agentConfig.name, ctx.currentModel);
-	const primaryModel = externalRunner ? undefined : params.modelOverrideFromParent
-		? params.modelOverride
-		: resolveSubagentModelOverride(
-			params.modelOverride ?? agentConfig.model,
-			ctx.currentModel,
-			availableModels,
-			ctx.currentModelProvider,
-			{ scope: modelScopes },
-		);
+	const modelOrigin = resolveModelOrigin({
+		fromParent: params.modelOverrideFromParent,
+		storedOrigin: params.modelOrigin,
+		explicitModel: params.modelOverrideFromParent ? undefined : params.modelOverride,
+		agentModel: agentConfig.model,
+		parentModel: ctx.currentModel,
+	});
+	let primaryModel: string | undefined;
+	try {
+		primaryModel = externalRunner ? undefined : modelOrigin === "inherited"
+			? params.modelOverride ?? (ctx.currentModel ? `${ctx.currentModel.provider}/${ctx.currentModel.id}` : undefined)
+			: resolveSubagentModelOverride(
+				params.modelOverride ?? agentConfig.model,
+				ctx.currentModel,
+				availableModels,
+				ctx.currentModelProvider,
+				{ scope: modelScopes, source: modelOrigin === "explicit" ? "explicit" : "inherited" },
+			);
+	} catch (error) {
+		return formatAsyncStartError("single", error instanceof Error ? error.message : String(error));
+	}
 	const effectiveThinking = externalRunner ? undefined : params.thinkingOverride ?? agentConfig.thinking;
 	const model = externalRunner ? undefined : applyThinkingSuffix(primaryModel, effectiveThinking, params.thinkingOverride !== undefined);
 	const contextLimit = model ? findModelInfo(model, availableModels, agentConfig.modelProvider ?? ctx.currentModelProvider)?.contextWindow : undefined;
@@ -1633,18 +1649,17 @@ export function executeAsyncSingle(
 	const structuredOutput = params.structuredOutputSchema
 		? createStructuredOutputRuntime(params.structuredOutputSchema, path.join(asyncDir, "structured-output"), { captureAcceptanceReport: params.acceptance !== false })
 		: undefined;
-	const modelCandidates = externalRunner
-		? []
-		: buildModelCandidates(primaryModel, agentConfig.fallbackModels, availableModels, agentConfig.modelProvider ?? ctx.currentModelProvider, {
-			scope: modelScopes,
-			primaryModelFromParent: params.modelOverrideFromParent,
-		})
-			.flatMap((candidate) => {
+	let modelCandidates: string[] = [];
+	if (!externalRunner) {
+		try {
+			modelCandidates = buildModelCandidates(primaryModel, agentConfig.fallbackModels, availableModels, agentConfig.modelProvider ?? ctx.currentModelProvider, {
+				scope: modelScopes,
+				primaryModelFromParent: modelOrigin === "inherited",
+				origin: modelOrigin,
+			}).flatMap((candidate) => {
 				const resolved = applyThinkingSuffix(candidate, effectiveThinking, params.thinkingOverride !== undefined);
 				return resolved ? [resolved] : [];
 			});
-	if (!externalRunner) {
-		try {
 			for (const candidate of modelCandidates) assertThinkingWithinCeiling({ model: candidate, configThinking: effectiveThinking, ceiling: thinkingCeiling, agent: agentConfig.name, runId: id });
 		} catch (error) {
 			return formatAsyncStartError("single", error instanceof Error ? error.message : String(error));
@@ -1730,7 +1745,8 @@ export function executeAsyncSingle(
 		...(model ? { model } : {}),
 		...(params.fast ?? recoveryAgentConfig.fast ? { fast: params.fast ?? recoveryAgentConfig.fast } : {}),
 		...(recoveryAgentConfig.modelProvider ? { modelProvider: recoveryAgentConfig.modelProvider } : {}),
-		...(params.modelOverrideFromParent ? { modelOverrideFromParent: true } : {}),
+		...(modelOrigin === "inherited" ? { modelOverrideFromParent: true } : {}),
+		modelOrigin,
 		...(recoveryAgentConfig.fallbackModels ? { fallbackModels: [...recoveryAgentConfig.fallbackModels] } : {}),
 		...(effectiveThinking ? { thinking: resolveEffectiveThinking(model, effectiveThinking) } : {}),
 		...(thinkingCeiling ? { thinkingCeiling } : {}),
@@ -1799,7 +1815,7 @@ export function executeAsyncSingle(
 						thinking: resolveEffectiveThinking(model, effectiveThinking),
 						...(thinkingCeiling ? { thinkingCeiling } : {}),
 						modelCandidates,
-						...(params.modelOverrideFromParent ? { skipPrimaryModelVerification: true } : {}),
+						...(modelOrigin === "inherited" ? { skipPrimaryModelVerification: true } : {}),
 						...(availableModels && availableModels.length > 0 ? { modelVerificationRegistry: availableModels } : {}),
 						tools: agentConfig.tools,
 						allowNestedSubagents: agentConfig.allowNestedSubagents,

package/src/runs/background/async-job-tracker.ts

@@ -73,7 +73,22 @@ export function createAsyncJobTracker(pi: Pick<ExtensionAPI, "events">, state: S
 	const steeringNoticeSeen = new Map<string, number>();
 	const jobWatchers = new Map<string, { watchers: Map<string, fs.FSWatcher>; retryTimer?: ReturnType<typeof setTimeout> }>();
 	const refreshTimers = new Map<string, ReturnType<typeof setTimeout>>();
+	let widgetRerenderTimer: ReturnType<typeof setTimeout> | undefined;
 	const runningJobIds = new Set<string>();
+	const externalJobBridgeRuns = new Set<string>();
+	const externalJobBridgeEligibility = (steps: AsyncJobState["steps"]): "required" | "not-required" | "unknown" => {
+		if (!Array.isArray(steps)) return "unknown";
+		for (const step of steps) {
+			const runner = (step as { runner?: unknown } | null)?.runner;
+			if (runner === undefined) continue;
+			if (!runner || typeof runner !== "object" || Array.isArray(runner)) return "unknown";
+			const runnerType = (runner as { type?: unknown }).type;
+			if (typeof runnerType !== "string") return "unknown";
+			if (runnerType === "external-job") return "required";
+			if (runnerType !== "pi" && runnerType !== "external-cli") return "unknown";
+		}
+		return "not-required";
+	};
 	let rootWatcher: fs.FSWatcher | undefined;
 	let nextLivenessAt = Date.now() + livenessIntervalMs;
 	let nextWidgetAnimationAt = Date.now() + WIDGET_ANIMATION_INTERVAL_MS;
@@ -94,9 +109,22 @@ export function createAsyncJobTracker(pi: Pick<ExtensionAPI, "events">, state: S
 	const rerenderLastWidget = (jobs = Array.from(state.asyncJobs.values())) => {
 		withLastUiContext((ctx) => rerenderWidget(ctx, jobs));
 	};
+	const scheduleLastWidgetRerender = () => {
+		if (widgetRerenderTimer) return;
+		widgetRerenderTimer = setTimeout(() => {
+			widgetRerenderTimer = undefined;
+			rerenderLastWidget();
+		}, EVENT_REFRESH_DEBOUNCE_MS);
+		widgetRerenderTimer.unref?.();
+	};
 	const requestLastWidgetRender = () => {
 		if (options.widgetEnabled === false) return;
-		rerenderLastWidget();
+		withLastUiContext((ctx) => {
+			if (state.widgetsSuspended) return;
+			const requestRender = (ctx.ui as { requestRender?: () => void }).requestRender;
+			if (requestRender) requestRender.call(ctx.ui);
+			else renderWidget(ctx, Array.from(state.asyncJobs.values()));
+		});
 	};
 	const refreshWidget = (ctx: ExtensionContext) => rerenderWidget(ctx);
 	const restoredControlEventCursor = (asyncDir: string) => {
@@ -340,6 +368,7 @@ export function createAsyncJobTracker(pi: Pick<ExtensionAPI, "events">, state: S
 		if (timer) clearTimeout(timer);
 		refreshTimers.delete(asyncId);
 		runningJobIds.delete(asyncId);
+		externalJobBridgeRuns.delete(asyncId);
 	};
 
 	const refreshJob = (job: AsyncJobState): boolean => {
@@ -354,9 +383,15 @@ export function createAsyncJobTracker(pi: Pick<ExtensionAPI, "events">, state: S
 				console.error(`Failed to refresh nested async descendants for '${job.asyncDir}':`, error);
 			}
 		};
+		let bridgeSweepAttempted = false;
 		try {
 			emitNewControlEvents(job);
-			serviceExternalJobBridgeRequests(job.asyncDir);
+			const bridgeAlreadyRequired = externalJobBridgeRuns.has(job.asyncId) || externalJobBridgeEligibility(job.steps) === "required";
+			if (bridgeAlreadyRequired) {
+				externalJobBridgeRuns.add(job.asyncId);
+				bridgeSweepAttempted = true;
+				serviceExternalJobBridgeRequests(job.asyncDir);
+			}
 			try {
 				if (job.nestedRoute) reconcileNestedAsyncDescendants(job.nestedRoute, { resultsDir, kill: options.kill, now: options.now });
 			} catch (error) {
@@ -382,6 +417,15 @@ export function createAsyncJobTracker(pi: Pick<ExtensionAPI, "events">, state: S
 				},
 			});
 			const status = reconciliation.status ?? readStatus(job.asyncDir);
+			if (!bridgeAlreadyRequired) {
+				const bridgeEligibility = externalJobBridgeEligibility(status?.steps);
+				if (bridgeEligibility === "required") externalJobBridgeRuns.add(job.asyncId);
+				// Missing or ambiguous status retains the legacy sweep for recovery races.
+				if (bridgeEligibility !== "not-required") {
+					bridgeSweepAttempted = true;
+					serviceExternalJobBridgeRequests(job.asyncDir);
+				}
+			}
 			if (status) {
 				const previousStatus = job.status;
 				job.status = status.state;
@@ -456,6 +500,14 @@ export function createAsyncJobTracker(pi: Pick<ExtensionAPI, "events">, state: S
 				runningJobIds.add(job.asyncId);
 			}
 		} catch (error) {
+			if (!bridgeSweepAttempted) {
+				try {
+					bridgeSweepAttempted = true;
+					serviceExternalJobBridgeRequests(job.asyncDir);
+				} catch (bridgeError) {
+					console.error(`Failed to service external job bridge for '${job.asyncDir}':`, bridgeError);
+				}
+			}
 			if (job.status !== "failed") {
 				console.error(`Failed to read async status for '${job.asyncDir}':`, error);
 				job.status = "failed";
@@ -473,7 +525,7 @@ export function createAsyncJobTracker(pi: Pick<ExtensionAPI, "events">, state: S
 		const timer = setTimeout(() => {
 			refreshTimers.delete(asyncId);
 			const job = state.asyncJobs.get(asyncId);
-			if (job && refreshJob(job)) rerenderLastWidget();
+			if (job && refreshJob(job)) scheduleLastWidgetRerender();
 		}, delayMs);
 		timer.unref?.();
 		refreshTimers.set(asyncId, timer);
@@ -604,6 +656,7 @@ export function createAsyncJobTracker(pi: Pick<ExtensionAPI, "events">, state: S
 			: rawAgents;
 		const sessionRoot = state.liveAsyncSessionRoots?.get(info.id);
 		state.liveAsyncSessionRoots?.delete(info.id);
+		externalJobBridgeRuns.delete(info.id);
 		state.asyncJobs.set(info.id, {
 			asyncId: info.id,
 			asyncDir,
@@ -677,11 +730,15 @@ export function createAsyncJobTracker(pi: Pick<ExtensionAPI, "events">, state: S
 		for (const asyncId of jobWatchers.keys()) closeJobWatcher(asyncId);
 		for (const timer of refreshTimers.values()) clearTimeout(timer);
 		refreshTimers.clear();
+		if (widgetRerenderTimer) clearTimeout(widgetRerenderTimer);
+		widgetRerenderTimer = undefined;
 		runningJobIds.clear();
+		externalJobBridgeRuns.clear();
 	};
 
 	const resetJobs = (ctx?: ExtensionContext) => {
 		dispose();
+		state.statusProjectionSessionId = null;
 		for (const timer of state.cleanupTimers.values()) clearTimeout(timer);
 		state.cleanupTimers.clear();
 		state.asyncJobs.clear();
@@ -698,6 +755,7 @@ export function createAsyncJobTracker(pi: Pick<ExtensionAPI, "events">, state: S
 
 	const restoreActiveJobs = (ctx?: ExtensionContext) => {
 		if (ctx?.hasUI) state.lastUiContext = ctx;
+		state.statusProjectionSessionId = null;
 		if (!state.currentSessionId) return;
 		let runs: AsyncRunSummary[];
 		try {
@@ -713,6 +771,7 @@ export function createAsyncJobTracker(pi: Pick<ExtensionAPI, "events">, state: S
 			rememberFleetJob(state, job);
 			watchJob(job);
 		}
+		state.statusProjectionSessionId = state.currentSessionId;
 		if (runs.length === 0) return;
 		ensurePoller();
 		rerenderLastWidget();

package/src/runs/background/async-resume.ts

@@ -314,7 +314,7 @@ export function readAsyncRecoveryDescriptor(asyncDir: string | undefined): Steer
 	if (!value || typeof value !== "object" || Array.isArray(value)) throw new Error(`Invalid async recovery descriptor '${descriptorPath}': expected an object.`);
 	const parsed = value as Record<string, unknown>;
 	const allowedFields = new Set([
-		"version", "launchContractDigest", "sourceRunId", "agentContract", "agent", "sessionFile", "cwd", "model", "modelProvider", "modelOverrideFromParent", "fallbackModels", "thinking", "thinkingCeiling", "tools", "allowNestedSubagents", "extensions",
+		"version", "launchContractDigest", "sourceRunId", "agentContract", "agent", "sessionFile", "cwd", "model", "modelProvider", "modelOverrideFromParent", "modelOrigin", "fallbackModels", "thinking", "thinkingCeiling", "tools", "allowNestedSubagents", "extensions",
 		"subagentOnlyExtensions", "mcpDirectTools", "mutationTools", "systemPrompt", "systemPromptMode", "inheritProjectContext", "inheritGlobalContext", "inheritSkills", "skills",
 		"skillPath", "agentFilePath", "completionGuard", "memory", "outputPath", "outputMode", "structuredOutputSchema", "acceptance", "sessionDir", "artifactConfig",
 		"artifactsDir", "maxOutput", "controlConfig", "context", "intercomBridge", "absoluteDeadlineAt", "initialTurnBudget", "initialToolBudget", "maxSubagentDepth", "share", "capabilityCeiling",
@@ -347,6 +347,8 @@ export function readAsyncRecoveryDescriptor(asyncDir: string | undefined): Steer
 	if (parsed.outputMode !== "inline" && parsed.outputMode !== "file-only") throw new Error(`Invalid async recovery descriptor '${descriptorPath}': outputMode is invalid.`);
 	if (parsed.context !== undefined && parsed.context !== "fresh" && parsed.context !== "fork") throw new Error(`Invalid async recovery descriptor '${descriptorPath}': context is invalid.`);
 	if (parsed.modelOverrideFromParent !== undefined && typeof parsed.modelOverrideFromParent !== "boolean") throw new Error(`Invalid async recovery descriptor '${descriptorPath}': modelOverrideFromParent must be a boolean.`);
+	if (parsed.modelOrigin !== undefined && parsed.modelOrigin !== "explicit" && parsed.modelOrigin !== "inherited" && parsed.modelOrigin !== "configured") throw new Error(`Invalid async recovery descriptor '${descriptorPath}': modelOrigin must be 'explicit', 'inherited', or 'configured'.`);
+	if (parsed.modelOrigin === undefined && parsed.model !== undefined) parsed.modelOrigin = parsed.modelOverrideFromParent ? "inherited" : "configured";
 	for (const field of ["inheritProjectContext", "inheritSkills", "share"] as const) {
 		if (typeof parsed[field] !== "boolean") throw new Error(`Invalid async recovery descriptor '${descriptorPath}': ${field} must be a boolean.`);
 	}

package/src/runs/background/async-status.ts

@@ -166,6 +166,40 @@ function getErrorMessage(error: unknown): string {
 	return error instanceof Error ? error.message : String(error);
 }
 
+function isAsyncStatusIsolationError(asyncDir: string, error: unknown): boolean {
+	const statusPath = path.join(asyncDir, "status.json");
+	const message = getErrorMessage(error);
+	return /^(?:Failed to (?:inspect|read|parse|validate) async status file|Invalid async status file) '/.test(message)
+		|| message.startsWith(`Invalid async status '${statusPath}'`)
+		|| message.startsWith(`Invalid host step '${statusPath}`)
+		|| /^(workflowChildren|Invalid workflowChildren)/.test(message);
+}
+
+function isolateCorruptActiveRun(asyncDir: string, runId: string, error: unknown, now?: () => number): void {
+	const statusPath = path.join(asyncDir, "status.json");
+	const processTerminal = readProcessTerminal(asyncDir, { runId });
+	let markerAge: number | undefined;
+	try {
+		markerAge = activeRunMarkerAgeMs(asyncDir, now?.());
+	} catch (markerError) {
+		console.error(`Failed to inspect corrupt async active-run marker for '${runId}':`, markerError);
+	}
+	const markerCanBeReleased = processTerminal?.state === "observed"
+		|| (markerAge !== undefined && markerAge > DEFAULT_STALE_TERMINAL_ACTIVE_MARKER_MS);
+	let markerAction = "active marker retained because runner liveness is unknown";
+	if (markerCanBeReleased) {
+		try {
+			releaseActiveRunIndex(asyncDir);
+			markerAction = processTerminal?.state === "observed"
+				? "active marker released after observed process-terminal proof"
+				: "stale active marker released";
+		} catch (releaseError) {
+			markerAction = `failed to release active marker: ${getErrorMessage(releaseError)}`;
+		}
+	}
+	console.error(`[pi-subagents] Skipping corrupt active async run '${runId}' at '${statusPath}': ${getErrorMessage(error)}; ${markerAction}.`);
+}
+
 function isNotFoundError(error: unknown): boolean {
 	return typeof error === "object"
 		&& error !== null
@@ -250,14 +284,16 @@ function deriveAsyncActivityState(asyncDir: string, status: AsyncStatus): { acti
 }
 
 function statusToSummary(asyncDir: string, status: AsyncStatus & { cwd?: string }, nestedWarnings: string[] = [], nestedRoute?: NestedRoute): AsyncRunSummary {
-	validateAsyncStatusLaneMetadata(status, `Invalid async status '${path.join(asyncDir, "status.json")}'`);
+	const statusPath = path.join(asyncDir, "status.json");
+	validateAsyncStatusLaneMetadata(status, `Invalid async status '${statusPath}'`);
 	const workflowChildren = parseWorkflowChildSummary(status.workflowChildren);
-	if (workflowChildren && workflowChildren.workflowRunId !== status.runId) throw new Error(`Invalid async status '${path.join(asyncDir, "status.json")}': workflowChildren.workflowRunId does not match.`);
-	assertWorkflowGraphHostSteps(status.workflowGraph, path.join(asyncDir, "status.json"), status.runId);
+	if (workflowChildren && workflowChildren.workflowRunId !== status.runId) throw new Error(`Invalid async status '${statusPath}': workflowChildren.workflowRunId does not match.`);
+	assertWorkflowGraphHostSteps(status.workflowGraph, statusPath, status.runId);
 	const hostSteps = validHostStepNodes(status.workflowGraph);
 	if (status.sessionId !== undefined && typeof status.sessionId !== "string") {
-		throw new Error(`Invalid async status '${path.join(asyncDir, "status.json")}': sessionId must be a string.`);
+		throw new Error(`Invalid async status '${statusPath}': sessionId must be a string.`);
 	}
+	if (status.outputFile !== undefined && typeof status.outputFile !== "string") throw new Error(`Invalid async status '${statusPath}': outputFile must be a string.`);
 	const { activityState, lastActivityAt } = deriveAsyncActivityState(asyncDir, status);
 	const processTerminal = readProcessTerminal(asyncDir, { runId: status.runId, runnerProcessInstanceId: status.processTerminal?.runnerProcessInstanceId })
 		?? sanitizeProcessTerminal(status.processTerminal, { runId: status.runId, runnerProcessInstanceId: status.processTerminal?.runnerProcessInstanceId }, path.join(asyncDir, "status.json"));
@@ -499,10 +535,17 @@ export function listAsyncRuns(asyncDirRoot: string, options: AsyncRunListOptions
 	};
 	for (const entry of entries) {
 		const asyncDir = path.join(asyncDirRoot, entry);
-		const reconciliation = options.reconcile === false
-			? undefined
-			: reconcileAsyncRun(asyncDir, { resultsDir: options.resultsDir, kill: options.kill, now: options.now });
-		const status = (reconciliation?.status ?? readStatus(asyncDir)) as (AsyncStatus & { cwd?: string }) | null;
+		let status: (AsyncStatus & { cwd?: string }) | null;
+		try {
+			const reconciliation = options.reconcile === false
+				? undefined
+				: reconcileAsyncRun(asyncDir, { resultsDir: options.resultsDir, kill: options.kill, now: options.now });
+			status = (reconciliation?.status ?? readStatus(asyncDir)) as (AsyncStatus & { cwd?: string }) | null;
+		} catch (error) {
+			if (!activeEntries.has(entry) || !isAsyncStatusIsolationError(asyncDir, error)) throw error;
+			isolateCorruptActiveRun(asyncDir, entry, error, options.now);
+			continue;
+		}
 		if (!status) {
 			if (activeEntries.has(entry)) updateActiveRunIndex(asyncDir, "failed");
 			continue;
@@ -528,7 +571,14 @@ export function listAsyncRuns(asyncDirRoot: string, options: AsyncRunListOptions
 				nestedWarnings.push(`Nested status unavailable: ${getErrorMessage(error)}`);
 			}
 		}
-		const summary = statusToSummary(asyncDir, status, nestedWarnings, nestedRoute);
+		let summary: AsyncRunSummary;
+		try {
+			summary = statusToSummary(asyncDir, status, nestedWarnings, nestedRoute);
+		} catch (error) {
+			if (!activeEntries.has(entry) || !isAsyncStatusIsolationError(asyncDir, error)) throw error;
+			isolateCorruptActiveRun(asyncDir, entry, error, options.now);
+			continue;
+		}
 		runs.push(summary);
 	}
 

package/src/runs/background/auto-drain.ts

@@ -62,7 +62,7 @@ export async function drainOutstandingWork(deps: AutoDrainDeps): Promise<void> {
 			},
 		);
 		if (waitResult.isError) {
-			throw new Error(`Auto-drain failed for session '${sessionId}': ${resultText(waitResult) || "subagent_wait returned an error without details"}.`);
+			throw new Error(`Auto-drain failed for session '${sessionId}': ${resultText(waitResult) || "bg_wait returned an error without details"}.`);
 		}
 	}
 }

package/src/runs/background/fleet-view.ts

@@ -330,7 +330,7 @@ function formatDetachedForegroundFleetLines(runs: ForegroundRun[]): string[] {
 		const childSummary = detachedChildren.map((child) => `${fleetChildDisplayName(child)} #${child.index}`).join(", ");
 		lines.push(`- ${run.runId} | detached | ${run.mode}${childSummary ? ` | ${childSummary}` : ""}`);
 		lines.push(`  status: subagent({ action: "status", id: "${run.runId}" })`);
-		lines.push(`  recovery: reply to the supervisor request first, then wait with subagent_wait({ id: "${run.runId}" }); do not resume or launch a replacement while any child remains detached.`);
+		lines.push(`  recovery: reply to the supervisor request first, then wait with bg_wait({ id: "${run.runId}" }); do not resume or launch a replacement while any child remains detached.`);
 	}
 	return lines;
 }

package/src/runs/background/result-watcher.ts

@@ -426,7 +426,7 @@ export function createResultWatcher(
 			if (observerSucceeded) removeMissionObserverIndex(resultsDir, runId);
 			const epoch = deliveryEpoch;
 			if (!ownsCompletion(sessionId, completionOwnerId, epoch)) return;
-			// Recorded before dedupe and before the unlink below so subagent_wait can
+			// Recorded before dedupe and before the unlink below so bg_wait can
 			// use the in-memory record or its bounded durable replay after cleanup.
 			recordWaitCompletion(state, runId, data, Date.now(), completionTtlMs, {
 				resultsDir,

package/src/runs/background/resume-guidance.ts

@@ -12,7 +12,7 @@ function isIntercomDetached(run: AsyncRunSummary): boolean {
 
 function formatIntercomDetachGuidance(run: AsyncRunSummary): string | undefined {
 	if (!isIntercomDetached(run)) return undefined;
-	return `Run "${run.id}" detached for intercom coordination. Reply to the supervisor request first, then wait with subagent_wait({ id: "${run.id}" }). Use subagent({ action: "status", id: "${run.id}" }) to recover the result; do not resume or launch a replacement while it remains detached.`;
+	return `Run "${run.id}" detached for intercom coordination. Reply to the supervisor request first, then wait with bg_wait({ id: "${run.id}" }). Use subagent({ action: "status", id: "${run.id}" }) to recover the result; do not resume or launch a replacement while it remains detached.`;
 }
 
 export function formatAsyncReviveCommand(run: AsyncRunSummary): string | undefined {

package/src/runs/background/run-status.ts

@@ -128,7 +128,7 @@ function formatResumeGuidance(runId: string | undefined, children: Array<{ agent
 	const resumableWorkflowChildren = workflowChildren.filter(({ child }) => !(child.status === "paused" && child.activityState === "needs_attention"));
 	if (workflowChildren.length > 0) {
 		return [
-			...supervisorDetachedWorkflowChildren.map(({ child }) => `Recovery workflow child${typeof child.workflowKey === "string" && child.workflowKey.trim() ? ` '${child.workflowKey}'` : ""}: reply to the supervisor request first, then wait with subagent_wait({ id: "${child.runId}" }). Use subagent({ action: "status", id: "${child.runId}" }) to recover the result; do not resume or launch a replacement while it remains detached.`),
+			...supervisorDetachedWorkflowChildren.map(({ child }) => `Recovery workflow child${typeof child.workflowKey === "string" && child.workflowKey.trim() ? ` '${child.workflowKey}'` : ""}: reply to the supervisor request first, then wait with bg_wait({ id: "${child.runId}" }). Use subagent({ action: "status", id: "${child.runId}" }) to recover the result; do not resume or launch a replacement while it remains detached.`),
 			...resumableWorkflowChildren.map(({ child }) => `Revive workflow child${typeof child.workflowKey === "string" && child.workflowKey.trim() ? ` '${child.workflowKey}'` : ""}: subagent({ action: "resume", id: "${child.runId}", message: "..." })`),
 		].join("\n");
 	}
@@ -228,7 +228,7 @@ function formatRememberedForegroundStatus(run: ForegroundResumeRun): string {
 	const detached = run.children.some((child) => child.status === "detached");
 	const resumable = run.children.find((child) => hasExistingSessionFile(child.sessionFile));
 	if (detached) {
-		lines.push(`Recovery: reply to the supervisor request first, then wait with subagent_wait({ id: "${run.runId}" }); do not resume or launch a replacement while any child remains detached.`);
+		lines.push(`Recovery: reply to the supervisor request first, then wait with bg_wait({ id: "${run.runId}" }); do not resume or launch a replacement while any child remains detached.`);
 	} else if (resumable) {
 		lines.push(run.children.length === 1
 			? `Revive: subagent({ action: "resume", id: "${run.runId}", message: "..." })`

package/src/runs/background/subagent-runner.ts

@@ -2149,7 +2149,7 @@ async function runSingleStepInner(
 			report: structuredAcceptanceReport as import("../../shared/types.ts").AcceptanceReport | undefined,
 			reportError: structuredAcceptanceReportError,
 			fileOutput: childWrittenOutput !== undefined && step.outputPath
-				? { content: childWrittenOutput, path: step.outputPath, authoritative: step.outputMode === "file-only" }
+				? { content: childWrittenOutput, path: step.outputPath, authoritative: step.outputMode === "file-only", durable: resolvedOutput.savedPath !== undefined }
 				: undefined,
 			cwd: step.cwd ?? ctx.cwd,
 			signal: combinedAbortSignal([ctx.timeoutSignal, ctx.stopSignal]),
@@ -5347,7 +5347,6 @@ async function runSubagent(
 			statusPayload.error = `Step failed: ${failedStep.agent}`;
 		}
 	}
-	writeStatusPayload();
 	try {
 		runPersistence.write(resultPath, {
 			lifecycleArtifactVersion: SUBAGENT_LIFECYCLE_ARTIFACT_VERSION,