@henryqw/pi-subagent 22.2.1 → 22.2.3

package/README.md

@@ -126,7 +126,7 @@ An isolated request has one durable ID, one goal, and 1–8 typed tasks:
 
 `dependsOn` controls scheduling. `contextFrom` may name completed text tasks and preserves the declared order. Changesets require focused task checks; a graph with changesets requires final checks. A canonical root `package.json` test script is required: the runner adds `pnpm test` to final checks if missing, and Main runs it on the combined integration checkout before promotion. Checks and judgments bind to exact Git identities; candidate drift, Main drift, mutation during validation, ambiguity, or conflicts stop promotion and retain evidence.
 
-`delegate_task`, `subagent_resume`, and `subagent_integrate advance` return a durable request ID after the wave is saved; productive work continues while Main is free. Pi sends a follow-up in the launching session whenever a checked worker becomes ready, and again when the wave finishes or needs attention. Use `subagent_status` if delivery is missed or you need current state.
+`delegate_task`, `subagent_resume`, and `subagent_integrate advance` return a durable request ID after the wave is saved; productive work continues while Main is free. Main can do other work or end its turn instead of sleeping or polling. Pi sends a follow-up in the launching session whenever a checked worker becomes ready, and again when the wave finishes or needs attention. Use `subagent_status` if delivery is missed or you need exact current state.
 
 The worker produces a committed candidate without tracked or untracked changes, then runs focused preliminary checks and optional judgment. Ignored generated files do not block candidacy, but prevent automatic worktree cleanup; inspect them before deleting retained work. The runner records each ready candidate immediately but **does not integrate it automatically**. Main can inspect `subagent_status` and use `subagent_stage` to select the exact candidate in an owned integration worktree while siblings are still working; conflicts are resolved there. Stage and resolve are available during an active wave, but rejection and revision wait for the wave to settle. `subagent_integrate` can advance dependent tasks from a staged snapshot, validate the combined tip with full root checks and optional final judgment after the wave settles, and promote it to Main only after exact validation. A definitive failed combined check/review allows one committed correction in the integration checkout followed by another validation. Main must explicitly promote; promotion and cleanup never silently discard retained work.
 

package/dist/herdr-runtime.js

@@ -152,9 +152,6 @@ function worktreeIntent(attempt) {
 export function workspaceLabel(attempt) {
     return attempt.correlationToken.slice(-6);
 }
-function legacyWorkspaceLabel(requestId, task, attempt) {
-    return `${requestId}/${task.id}#${attempt.number}`;
-}
 function expectedWorkerLabel(task) {
     return `${task.role}/${task.modelClass}`;
 }
@@ -174,10 +171,9 @@ function requireIntentIdentity(intent, attempt) {
         throw new Error(`${intent.kind} allocation intent does not match the task attempt.`);
     }
 }
-function assertWorkspaceIntent(intent, requestId, task, attempt) {
+function assertWorkspaceIntent(intent, attempt) {
     const worktree = worktreeIntent(attempt);
-    const label = exactString(intent.label, "workspace label");
-    if ((label !== workspaceLabel(attempt) && label !== legacyWorkspaceLabel(requestId, task, attempt))
+    if (exactString(intent.label, "workspace label") !== workspaceLabel(attempt)
         || exactAbsolutePath(intent.worktreeCwd, "workspace worktree cwd") !== worktree.cwd
         || exactAbsolutePath(intent.mainRoot, "workspace Main root") !== worktree.repoRoot) {
         throw new Error("Workspace allocation plan drifted from the exact owned worktree.");
@@ -368,7 +364,7 @@ export class HerdrHostRuntime {
         if (input.intent.kind === "workspace") {
             if (input.acquireLaunch)
                 throw new Error("Role launch acquisition is valid only at the exact agent allocation boundary.");
-            return await this.allocateWorkspace(input.intent, input.requestId, input.task, input.attempt, context);
+            return await this.allocateWorkspace(input.intent, input.attempt, context);
         }
         if (input.intent.kind === "worker_tab") {
             if (input.acquireLaunch)
@@ -381,7 +377,7 @@ export class HerdrHostRuntime {
         requireIntentIdentity(input.intent, input.attempt);
         const allocation = input.intent;
         if (allocation.kind === "workspace") {
-            assertWorkspaceIntent(allocation, input.requestId, input.task, input.attempt);
+            assertWorkspaceIntent(allocation, input.attempt);
             await this.assertRepositoryIdentity(allocation, context);
             const response = await this.herdr.json(["worktree", "list", "--cwd", allocation.worktreeCwd], this.processOptions(allocation.herdrRepoRoot, context, HERDR_OPERATION_CAP_MS));
             const result = resultRecord(response, "Herdr worktree list response");
@@ -656,7 +652,7 @@ export class HerdrHostRuntime {
                 assertWorkerTabIntent(allocation, input.task, input.attempt);
                 const leasePath = exactString(allocation.leasePath, "saved worker lease path");
                 this.assertLeasePath(leasePath, input.attempt.correlationToken);
-                assertWorkspaceIntent(workspace, input.requestId, input.task, input.attempt);
+                assertWorkspaceIntent(workspace, input.attempt);
                 await this.assertRepositoryIdentity(workspace, context);
                 const tabId = exactString(allocation.tabId, "saved worker tab ID");
                 const workspaceId = exactString(workspace.workspaceId, "saved workspace ID");
@@ -688,7 +684,7 @@ export class HerdrHostRuntime {
                 return { outcome: "blocked", failure: "Workspace cleanup must follow worker-tab reconciliation." };
             const allocation = ownedIntent(input.attempt, "workspace");
             requireIntentIdentity(allocation, input.attempt);
-            assertWorkspaceIntent(allocation, input.requestId, input.task, input.attempt);
+            assertWorkspaceIntent(allocation, input.attempt);
             await this.assertRepositoryIdentity(allocation, context);
             const workspaceId = exactString(allocation.workspaceId, "saved workspace ID");
             if (await this.workspaceAbsent(workspaceId, allocation, context))
@@ -708,8 +704,8 @@ export class HerdrHostRuntime {
             return { outcome: "blocked", failure: safeText(error) };
         }
     }
-    async allocateWorkspace(allocation, requestId, task, attempt, context) {
-        assertWorkspaceIntent(allocation, requestId, task, attempt);
+    async allocateWorkspace(allocation, attempt, context) {
+        assertWorkspaceIntent(allocation, attempt);
         await this.assertRepositoryIdentity(allocation, context);
         const args = ["worktree", "open", "--cwd", allocation.herdrRepoRoot, "--path", allocation.worktreeCwd, "--label", allocation.label, "--no-focus"];
         const response = await this.herdr.exec(args, this.processOptions(allocation.herdrRepoRoot, context, HERDR_OPERATION_CAP_MS));

package/extensions/result-transport.ts

@@ -3,7 +3,7 @@ import type { WorkflowEntry, WorkflowMode } from "./workflow.ts";
 
 const EVIDENCE_PREVIEW_CODE_POINTS = 256;
 
-export type WorkflowTransportStatus = "pending" | "running" | "succeeded" | "failed" | "rejected" | "skipped";
+export type WorkflowTransportStatus = "pending" | "running" | "succeeded" | "rejected" | "skipped";
 
 type TransportEntryBase = {
 	id: WorkflowEntry["id"];
@@ -17,7 +17,7 @@ type TransportEntryBase = {
 export type WorkflowTransportEntry =
 	| TransportEntryBase & { status: "pending" | "skipped"; assistantOutput?: never; failure?: never }
 	| TransportEntryBase & { status: "running" | "succeeded"; assistantOutput: string; failure?: never }
-	| TransportEntryBase & { status: "failed" | "rejected"; assistantOutput?: never; failure: string };
+	| TransportEntryBase & { status: "rejected"; assistantOutput?: never; failure: string };
 
 export type WorkflowTransportEntryDetails = {
 	id: string;
@@ -37,7 +37,7 @@ export type WorkflowTransportDetails = {
 
 export type BackgroundWorkflowTransportDetails = WorkflowTransportDetails & {
 	taskId: string;
-	outcome: "completed" | "failed" | "aborted";
+	outcome: "completed" | "failed";
 	recovery?: true;
 };
 
@@ -63,7 +63,7 @@ function workflowTitle(mode: WorkflowMode): string {
 function statusCounts(entries: readonly WorkflowTransportEntry[]): string[] {
 	const count = (statuses: WorkflowTransportStatus[]) => entries.filter(({ status }) => statuses.includes(status)).length;
 	return ([
-		[["failed", "rejected"], "failed"],
+		[["rejected"], "failed"],
 		[["succeeded"], "completed"],
 		[["skipped"], "skipped"],
 	] as const).flatMap(([statuses, word]) => {
@@ -84,7 +84,6 @@ const ENTRY_STATUS_PRESENTATION = {
 	pending: { glyph: "○", fallback: "queued" },
 	running: { glyph: "◌", fallback: "working" },
 	succeeded: { glyph: "✓", fallback: "completed" },
-	failed: { glyph: "✗", fallback: "failed" },
 	rejected: { glyph: "✗", fallback: "failed" },
 	skipped: { glyph: "–", fallback: "skipped" },
 } as const satisfies Record<WorkflowTransportEntryDetails["status"], WorkflowEntryStatusPresentation>;
@@ -94,7 +93,7 @@ export function presentWorkflowEntryStatus(status: WorkflowTransportEntryDetails
 }
 
 function sourceFor(entry: WorkflowTransportEntry): string | undefined {
-	if (entry.status === "failed" || entry.status === "rejected") return entry.failure;
+	if (entry.status === "rejected") return entry.failure;
 	if (entry.status === "running" || entry.status === "succeeded") return entry.assistantOutput;
 }
 
@@ -103,7 +102,7 @@ function evidenceFor(entry: WorkflowTransportEntry, position: number, total: num
 	if (source === undefined || !source) return;
 	const [preview, remainder] = splitEvidence(source);
 	return {
-		heading: `- [${position}/${total}] ${entry.name} · ${entry.role} · ${entry.status === "failed" || entry.status === "rejected" ? "failure" : "result"}:`,
+		heading: `- [${position}/${total}] ${entry.name} · ${entry.role} · ${entry.status === "rejected" ? "failure" : "result"}:`,
 		preview,
 		remainder,
 	};
@@ -114,7 +113,7 @@ export function formatWorkflowResult(mode: WorkflowMode, entries: readonly Workf
 	if (ordered.some(({ status }) => status === "pending" || status === "running")) {
 		throw new TypeError("Final workflow transport requires terminal entry states.");
 	}
-	const failed = ordered.some(({ status }) => status === "failed" || status === "rejected");
+	const failed = ordered.some(({ status }) => status === "rejected");
 	const positioned = ordered.map((entry, index) => ({ entry, position: index + 1 }));
 	const evidence = positioned.flatMap(({ entry, position }) => {
 		const value = evidenceFor(entry, position, ordered.length);

package/extensions/subagent.ts

@@ -2,7 +2,7 @@ import { randomUUID } from "node:crypto";
 import type { ExtensionAPI, ExtensionContext, Theme } from "@earendil-works/pi-coding-agent";
 import { type Component, Text, type TUI, truncateToWidth, visibleWidth } from "@earendil-works/pi-tui";
 import { availableTaskModels, loadTaskModelsConfig, modelReference, registerModelTask, resolveAvailableModel, type ResolvedTaskRoute, taskThinkingLevels } from "@henryqw/pi-task-models";
-import { capEphemeralSubagentOutput as capOutput, createEphemeralSubagentExecutor, DELEGATE_TASK, formatDuration, loadRoles, prepareRoleLaunch, ROLE_TOOL_POLICY_FLAG, type EphemeralSubagentTimeout, type Role } from "@henryqw/pi-subagent";
+import { capEphemeralSubagentOutput as capOutput, createEphemeralSubagentExecutor, DELEGATE_TASK, formatDuration, loadRoles, prepareRoleLaunch, ROLE_TOOL_POLICY_FLAG, type Role } from "@henryqw/pi-subagent";
 import { DEFAULT_EXECUTION_POLICY, DEFAULT_TIMEOUT_CONFIG, readSubagentConfig, resolveExecutionPolicy, type EffectiveExecutionPolicy } from "./config.ts";
 import { createCheckoutAdmission, roleCanWrite } from "./admission.ts";
 import { registerIsolatedExtension } from "./isolated.ts";
@@ -18,8 +18,7 @@ const MAX_WIDGET_ITEMS = 8;
 const MAX_WIDGET_LINES = 6;
 const SPINNER_FRAMES = ["⠋", "⠙", "⠹", "⠸", "⠼", "⠴", "⠦", "⠧", "⠇", "⠏"];
 
-type TimeoutPolicy = EphemeralSubagentTimeout;
-type WidgetStatus = "working" | "success" | "failure" | "aborted";
+type WidgetStatus = "working" | "success" | "failure";
 type WidgetItem = {
 	role: string;
 	model: string;
@@ -42,7 +41,6 @@ function statusGlyph(status: WidgetStatus, spinnerIndex: number, theme: Theme):
 		case "working": return theme.fg("accent", SPINNER_FRAMES[spinnerIndex % SPINNER_FRAMES.length]!);
 		case "success": return theme.fg("success", "✓");
 		case "failure": return theme.fg("error", "✗");
-		case "aborted": return theme.fg("warning", "■");
 	}
 }
 
@@ -51,7 +49,6 @@ function statusLabel(status: WidgetStatus): string {
 		case "working": return "working";
 		case "success": return "complete";
 		case "failure": return "failed";
-		case "aborted": return "stopped";
 	}
 }
 
@@ -87,11 +84,11 @@ function renderWidgetRows(
 	});
 	const hidden = ordered.slice(visible.length);
 	if (hidden.length) {
-		const counts: Record<WidgetStatus, number> = { working: 0, success: 0, failure: 0, aborted: 0 };
+		const counts: Record<WidgetStatus, number> = { working: 0, success: 0, failure: 0 };
 		for (const { status } of hidden) counts[status] += 1;
 		lines.push(truncateToWidth(theme.fg("muted", [
 			`… ${hidden.length} more`,
-			...(["working", "success", "failure", "aborted"] as const).flatMap((status) =>
+			...(["working", "success", "failure"] as const).flatMap((status) =>
 				counts[status] ? [`${counts[status]} ${statusLabel(status)}`] : []),
 		].join(" · ")), width));
 	}
@@ -128,10 +125,7 @@ const roleSummary = (): string => {
 	}
 };
 
-export default function subagentExtension(
-	pi: ExtensionAPI,
-	overrideTimeoutPolicy?: TimeoutPolicy,
-): void {
+export default function subagentExtension(pi: ExtensionAPI): void {
 	if (process.argv.includes(`--${ROLE_TOOL_POLICY_FLAG}`)) return;
 	registerModelTask(pi, DELEGATE_TASK);
 	pi.registerMessageRenderer(DIRECT_RESULT_TYPE, (message, { expanded, outputPad }, theme) => {
@@ -142,11 +136,9 @@ export default function subagentExtension(
 		if (!details?.entries) return new Text(content, outputPad, 0);
 		const count = details.entries.length;
 		const subject = count === 1 ? "Direct subagent" : `${count} direct subagents`;
-		const state = details.recovery ? "stopped; recovery needed"
-			: details.outcome === "completed" ? "completed"
-				: details.outcome === "failed" ? "failed" : "stopped";
-		const glyph = details.recovery || details.outcome === "aborted" ? "■" : details.outcome === "completed" ? "✓" : "✗";
-		const color = details.recovery || details.outcome === "aborted" ? "warning" : details.outcome === "completed" ? "success" : "error";
+		const state = details.recovery ? "stopped; recovery needed" : details.outcome;
+		const glyph = details.recovery ? "■" : details.outcome === "completed" ? "✓" : "✗";
+		const color = details.recovery ? "warning" : details.outcome === "completed" ? "success" : "error";
 		const rows = details.entries.map(({ name, role, status, summary }) => {
 			const { glyph, fallback } = presentWorkflowEntryStatus(status);
 			return `${glyph} ${name} · ${role} — ${details.recovery ? fallback : summary || fallback}`;
@@ -175,20 +167,11 @@ export default function subagentExtension(
 			maxCorrections: DEFAULT_EXECUTION_POLICY.maxCorrections,
 		};
 	}
-	const timeoutPolicy: TimeoutPolicy = overrideTimeoutPolicy ?? {
-		idleMs: initialPolicy.childIdleMs,
-		maxMs: initialPolicy.childMaxMs,
-	};
-	const effectiveInitialPolicy: EffectiveExecutionPolicy = {
-		...initialPolicy,
-		childIdleMs: timeoutPolicy.idleMs,
-		childMaxMs: timeoutPolicy.maxMs ?? initialPolicy.childMaxMs,
-	};
 	const executor = createEphemeralSubagentExecutor({
-		maxConcurrency: effectiveInitialPolicy.maxSubagents,
-		maxTurns: effectiveInitialPolicy.maxTurns,
-		maxTokens: effectiveInitialPolicy.maxTokens,
-		timeout: timeoutPolicy,
+		maxConcurrency: initialPolicy.maxSubagents,
+		maxTurns: initialPolicy.maxTurns,
+		maxTokens: initialPolicy.maxTokens,
+		timeout: { idleMs: initialPolicy.childIdleMs, maxMs: initialPolicy.childMaxMs },
 	});
 	const currentPolicy = (): EffectiveExecutionPolicy => resolveExecutionPolicy(readSubagentConfig());
 	const canWrite = (input: unknown): boolean => {
@@ -205,7 +188,7 @@ export default function subagentExtension(
 	admission.register(pi, canWrite);
 	const isolatedSurface = registerIsolatedExtension(pi, {
 		executor,
-		policy: effectiveInitialPolicy,
+		policy: initialPolicy,
 		currentPolicy,
 	});
 	let directSequence = 0;

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@henryqw/pi-subagent",
-  "version": "22.2.1",
+  "version": "22.2.3",
   "description": "Delegate bounded direct work or durable checked isolated task graphs to Pi roles.",
   "keywords": [
     "pi-package",

package/skills/pi-subagent/SKILL.md

@@ -18,6 +18,8 @@ Keep trivial, mechanically verifiable work in Main. Choose the user-requested mo
 
 Roles describe responsibility and capabilities; they do not select isolation. Bash, MCP tools, unknown tools, and arbitrary extensions are potentially writing. The checkout coordinator reduces races among Pi-owned calls but is not an OS sandbox and cannot control external processes.
 
+After delegation acknowledges a handle or durable request, do other work or end the turn; do not `sleep`, block on a Herdr CLI wait, or repeatedly poll for progress. Pi queues verified direct results and checked isolated candidates as follow-ups, which Main receives after its current turn. Use `subagent_status` for exact isolated evidence when a follow-up arrives or delivery was missed; Herdr lifecycle badges and screen output are not proof of completion.
+
 # Direct requests
 
 Use one compact `role`/`name`/`task` packet, `tasks` for independent packets, or `chain` for dependent packets. Only text tasks are accepted; a write-capable Role, `kind: changeset`, checks, and judgment require isolated mode. The tool returns a handle after launching Herdr tabs in the current workspace. Main receives the verified result as a follow-up after the workflow finishes, without interrupting an active turn. If observation stops, open `/subagent` to inspect exact recorded tabs and session files on this session branch; recorded does not prove still running.