@bastani/atomic 0.9.12 → 0.9.13-alpha.2

package/dist/builtin/subagents/src/tui/render-widget-graph.ts

@@ -1,14 +1,10 @@
 import { keyHintIfBound } from "@bastani/atomic";
-import { shortenPath } from "../shared/formatters.ts";
-import { aggregateStepStatus, formatParallelOutcome } from "../shared/status-format.ts";
 import type { AsyncJobState } from "../shared/types.ts";
-import { buildAsyncChainStepSpans } from "./render-chain-graph.ts";
 import {
 	formatNestedWidgetLines,
 	modelThinkingBadge,
 	widgetActivity,
 	widgetJobName,
-	widgetOutputPath,
 	widgetStats,
 	widgetStatusGlyph,
 	widgetStepActivity,
@@ -20,47 +16,6 @@ import {
 import { getTermWidth, type Theme, truncLine } from "./render-layout.ts";
 import { buildLiveStatusLine, themeBold } from "./render-status-progress.ts";
 
-export function widgetChainDetails(
-	job: AsyncJobState,
-	theme: Theme,
-	expanded = false,
-	width = getTermWidth(),
-	pulseFrame?: number,
-): string[] {
-	if (!job.steps?.length) return [];
-	const total = job.chainStepCount ?? job.steps.length;
-	const lines: string[] = [];
-	for (const span of buildAsyncChainStepSpans(total, job.steps.length, job.parallelGroups)) {
-		const steps = job.steps.slice(span.start, span.start + span.count);
-		if (span.isParallel) {
-			const status = aggregateStepStatus(steps);
-			lines.push(
-				`  ${widgetStepPulseGlyph(status, theme, pulseFrame)} Step ${span.stepIndex + 1}/${total}: ${themeBold(theme, "parallel group")} ${theme.fg("dim", "·")} ${theme.fg("dim", formatParallelOutcome(steps, span.count))}`,
-			);
-			continue;
-		}
-		const step = steps[0];
-		if (!step) {
-			lines.push(`  ${theme.fg("dim", `◦ Step ${span.stepIndex + 1}/${total}: pending`)}`);
-			continue;
-		}
-		lines.push(
-			...foregroundStyleWidgetStepLines(
-				job,
-				theme,
-				step,
-				"Step",
-				span.stepIndex + 1,
-				total,
-				expanded,
-				width,
-				pulseFrame,
-			),
-		);
-	}
-	return lines;
-}
-
 export function widgetParallelAgentDetails(
 	job: AsyncJobState,
 	theme: Theme,
@@ -70,15 +25,13 @@ export function widgetParallelAgentDetails(
 	pulseFrame?: number,
 ): string[] {
 	if (!job.steps?.length) return [];
-	if (job.mode !== "parallel" && job.mode !== "chain") return [];
-	if (job.mode === "chain" && !job.activeParallelGroup && job.parallelGroups?.length)
-		return widgetChainDetails(job, theme, expanded, width, pulseFrame);
+	if (job.mode !== "parallel") return [];
 	const total = job.stepsTotal ?? job.steps.length;
 	const lines: string[] = [];
 	for (const [index, step] of job.steps.entries()) {
 		const marker = index === job.steps.length - 1 ? "└" : "├";
 		const activity = widgetStepActivity(step, job.updatedAt);
-		const itemTitle = job.mode === "parallel" || job.activeParallelGroup ? "Agent" : "Step";
+		const itemTitle = "Agent";
 		const modelDisplay = modelThinkingBadge(theme, step.model, step.thinking, step.fastMode);
 		lines.push(
 			`  ${theme.fg("dim", `${marker} ${widgetStepPulseGlyph(step.status, theme, pulseFrame)} ${itemTitle} ${index + 1}/${total}: ${step.agent} · ${widgetStepStatus(step.status, theme)}${modelDisplay}${activity ? ` · ${activity}` : ""}`)}`,
@@ -130,8 +83,6 @@ export function foregroundStyleWidgetStepLines(
 	if (step.status === "running") {
 		const expandHint = keyHintIfBound("app.tools.expand", "for live detail");
 		if (!expanded && expandHint) lines.push(`    ${theme.fg("accent", `Press ${expandHint}`)}`);
-		const output = widgetOutputPath(job, step);
-		if (output) lines.push(`    ${theme.fg("dim", `output: ${shortenPath(output)}`)}`);
 		if (expanded) {
 			const liveStatus = buildLiveStatusLine(step, job.updatedAt);
 			if (liveStatus && liveStatus !== activity) lines.push(`    ${theme.fg("accent", liveStatus)}`);
@@ -168,10 +119,8 @@ export function foregroundStyleWidgetDetails(
 				pulseFrame,
 			).map((line) => `  ${line}`),
 		];
-	if (job.mode === "chain" && !job.activeParallelGroup && job.parallelGroups?.length)
-		return widgetChainDetails(job, theme, expanded, width, pulseFrame);
 	const total = job.stepsTotal ?? job.steps.length;
-	const itemTitle = job.mode === "parallel" || job.activeParallelGroup ? "Agent" : "Step";
+	const itemTitle = job.mode === "parallel" ? "Agent" : "Step";
 	const lines: string[] = [];
 	for (const [index, step] of job.steps.entries()) {
 		lines.push(
@@ -203,8 +152,7 @@ export function buildSingleWidgetLines(
 	pulseFrame?: number,
 ): string[] {
 	const stats = widgetStats(job, theme);
-	const count =
-		job.mode === "chain" ? job.chainStepCount : (job.stepsTotal ?? job.agents?.length ?? job.steps?.length);
+	const count = job.stepsTotal ?? job.agents?.length ?? job.steps?.length;
 	const mode = widgetJobName(job);
 	const title = `async subagent ${mode}${count && count > 1 ? ` (${count})` : ""}`;
 	return [
@@ -222,11 +170,10 @@ export function compactSingleWidgetLines(
 	pulseFrame?: number,
 ): string[] {
 	const fullLines = buildSingleWidgetLines(job, theme, width, false, _now, pulseFrame);
-	if (fullLines.length <= 10 || !job.steps?.length || (job.mode !== "parallel" && !job.activeParallelGroup))
-		return fullLines;
+	if (fullLines.length <= 10 || !job.steps?.length || job.mode !== "parallel") return fullLines;
 
 	const total = job.stepsTotal ?? job.steps.length;
-	const itemTitle = job.mode === "parallel" || job.activeParallelGroup ? "Agent" : "Step";
+	const itemTitle = "Agent";
 	const lines = fullLines.slice(0, 2);
 	for (const [index, step] of job.steps.entries()) {
 		const status = widgetStepStatus(step.status, theme);
@@ -247,10 +194,6 @@ export function compactSingleWidgetLines(
 			pulseFrame,
 		))
 			lines.push(`    ${nestedLine}`);
-		if (step.status === "running") {
-			const output = widgetOutputPath(job, step);
-			if (output) lines.push(`    ${theme.fg("dim", `output: ${shortenPath(output)}`)}`);
-		}
 	}
 	const expandHint = keyHintIfBound("app.tools.expand", "for live detail");
 	if (expandHint && job.steps.some((step) => step.status === "running"))

package/dist/builtin/subagents/src/tui/render-widget.ts

@@ -170,28 +170,6 @@ function getWidgetOwnerKey(ctx: ExtensionContext): string {
 	return `${sessionOwner}|cwd:${cwdOwner}`;
 }
 
-function widgetStatusText(jobs: AsyncJobState[]): string | undefined {
-	if (jobs.length === 0) return undefined;
-	const counts = new Map<string, number>();
-	for (const job of jobs) counts.set(job.status, (counts.get(job.status) ?? 0) + 1);
-	const ordered = ["running", "queued", "complete", "failed", "paused"];
-	const parts = ordered
-		.map((status) => {
-			const count = counts.get(status) ?? 0;
-			return count > 0 ? `${count} ${status}` : undefined;
-		})
-		.filter((part): part is string => part !== undefined);
-	return `Async agents: ${parts.join(", ")}`;
-}
-
-function setWidgetStatus(ctx: ExtensionContext, jobs: AsyncJobState[]): void {
-	try {
-		ctx.ui.setStatus?.(WIDGET_KEY, widgetStatusText(jobs));
-	} catch {
-		// Status mirroring must never prevent the primary widget render path.
-	}
-}
-
 function requestWidgetRender(ctx: ExtensionContext): void {
 	(ctx as RenderRequestingContext).ui.requestRender?.();
 }
@@ -343,7 +321,6 @@ export function renderWidget(ctx: ExtensionContext, jobs: AsyncJobState[], owner
 		if (requestedSurface.activeOwner && requestedSurface.activeOwner !== owner) return;
 		if (state.surfaceKey && getSurfaceState(state.surfaceKey).activeOwner !== owner) return;
 		if (state.mounted && state.mountedOwnerKey !== ownerKey) return;
-		if (ctx.hasUI) setWidgetStatus(ctx, []);
 		stopWidgetAnimation(ctx, owner);
 		return;
 	}
@@ -363,7 +340,6 @@ export function renderWidget(ctx: ExtensionContext, jobs: AsyncJobState[], owner
 		state.latestJobs = [...jobs];
 		refreshWidgetSnapshot(state, jobs);
 	}
-	setWidgetStatus(ctx, jobs);
 	if (!state.mounted) {
 		const surface = getSurfaceState(requestedSurfaceKey);
 		if (surface.activeOwner && surface.activeOwner !== owner) return;

package/dist/builtin/web-access/CHANGELOG.md

@@ -4,6 +4,20 @@ All notable changes to this project will be documented in this file.
 
 ## [Unreleased]
 
+## [0.9.13-alpha.2] - 2026-08-12
+
+### Changed
+
+- GitHub, credential-store, and Bun subprocesses now receive `AI_AGENT=atomic` for generic child-process attribution.
+
+### Fixed
+- Fixed background content fetch notifications that completed after an extension reload from throwing through stale runtime handles.
+- Fixed background fetch stale-context checks by consuming the host's exported predicate instead of copying its error-message marker.
+- Restored visible background content-fetch error notifications for non-stale success-path failures while keeping stale runtime errors contained and pending state alive until notification admission settles.
+
+- Routed query rewriting and summary-review model calls through Atomic's model runtime so registered custom providers and resolved request authentication apply.
+- Fixed frames and thumbnails returned by `fetch_content` bypassing the host's `images.autoResize` processing before they enter session history.
+
 ## [0.9.12] - 2026-08-04
 
 ### Fixed

package/dist/builtin/web-access/chrome-cookies.ts

@@ -3,6 +3,7 @@ import { pbkdf2Sync, createDecipheriv } from "node:crypto";
 import { copyFileSync, existsSync, mkdtempSync, rmSync } from "node:fs";
 import { tmpdir, homedir, platform } from "node:os";
 import { join } from "node:path";
+import { createChildProcessEnvironment } from "@bastani/atomic";
 
 export type CookieMap = Record<string, string>;
 
@@ -190,7 +191,7 @@ function readKeychainPassword(account: string, service: string): Promise<string
 		execFile(
 			"security",
 			["find-generic-password", "-w", "-a", account, "-s", service],
-			{ timeout: 5000 },
+			{ timeout: 5000, env: createChildProcessEnvironment() },
 			(err, stdout) => {
 				if (err) { resolve(null); return; }
 				resolve(stdout.trim() || null);
@@ -206,7 +207,7 @@ function readLinuxPassword(secretToolApp: string | undefined): Promise<string> {
 		execFile(
 			"secret-tool",
 			["lookup", "application", secretToolApp],
-			{ timeout: 5000 },
+			{ timeout: 5000, env: createChildProcessEnvironment() },
 			(err, stdout) => {
 				if (err) {
 					// KDE Wallet users fall through to peanuts intentionally.

package/dist/builtin/web-access/github-api.ts

@@ -1,4 +1,5 @@
 import { execFile } from "node:child_process";
+import { createChildProcessEnvironment } from "@bastani/atomic";
 import type { ExtractedContent } from "./extract.js";
 import { flattenTruncatedString } from "./flat-string.js";
 import type { GitHubUrlInfo } from "./github-extract.js";
@@ -13,7 +14,7 @@ export async function checkGhAvailable(): Promise<boolean> {
 	if (ghAvailable !== null) return ghAvailable;
 
 	return new Promise((resolve) => {
-		execFile("gh", ["--version"], { timeout: 5000 }, (err) => {
+		execFile("gh", ["--version"], { timeout: 5000, env: createChildProcessEnvironment() }, (err) => {
 			ghAvailable = !err;
 			resolve(ghAvailable);
 		});
@@ -31,7 +32,7 @@ export async function checkRepoSize(owner: string, repo: string): Promise<number
 	if (!(await checkGhAvailable())) return null;
 
 	return new Promise((resolve) => {
-		execFile("gh", ["api", `repos/${owner}/${repo}`, "--jq", ".size"], { timeout: 10000 }, (err, stdout) => {
+		execFile("gh", ["api", `repos/${owner}/${repo}`, "--jq", ".size"], { timeout: 10000, env: createChildProcessEnvironment() }, (err, stdout) => {
 			if (err) {
 				resolve(null);
 				return;
@@ -46,7 +47,7 @@ async function getDefaultBranch(owner: string, repo: string): Promise<string | n
 	if (!(await checkGhAvailable())) return null;
 
 	return new Promise((resolve) => {
-		execFile("gh", ["api", `repos/${owner}/${repo}`, "--jq", ".default_branch"], { timeout: 10000 }, (err, stdout) => {
+		execFile("gh", ["api", `repos/${owner}/${repo}`, "--jq", ".default_branch"], { timeout: 10000, env: createChildProcessEnvironment() }, (err, stdout) => {
 			if (err) {
 				resolve(null);
 				return;
@@ -64,7 +65,7 @@ async function fetchTreeViaApi(owner: string, repo: string, ref: string): Promis
 		execFile(
 			"gh",
 			["api", `repos/${owner}/${repo}/git/trees/${ref}?recursive=1`, "--jq", ".tree[].path"],
-			{ timeout: 15000, maxBuffer: 5 * 1024 * 1024 },
+			{ timeout: 15000, maxBuffer: 5 * 1024 * 1024, env: createChildProcessEnvironment() },
 			(err, stdout) => {
 				if (err) {
 					resolve(null);
@@ -90,7 +91,7 @@ async function fetchReadmeViaApi(owner: string, repo: string, ref: string): Prom
 		execFile(
 			"gh",
 			["api", `repos/${owner}/${repo}/readme?ref=${ref}`, "--jq", ".content"],
-			{ timeout: 10000 },
+			{ timeout: 10000, env: createChildProcessEnvironment() },
 			(err, stdout) => {
 				if (err) {
 					resolve(null);
@@ -114,7 +115,7 @@ async function fetchFileViaApi(owner: string, repo: string, path: string, ref: s
 		execFile(
 			"gh",
 			["api", `repos/${owner}/${repo}/contents/${path}?ref=${ref}`, "--jq", ".content"],
-			{ timeout: 10000, maxBuffer: 2 * 1024 * 1024 },
+			{ timeout: 10000, maxBuffer: 2 * 1024 * 1024, env: createChildProcessEnvironment() },
 			(err, stdout) => {
 				if (err) {
 					resolve(null);

package/dist/builtin/web-access/github-extract.ts

@@ -1,6 +1,7 @@
 import { existsSync, readFileSync, rmSync, statSync, readdirSync, openSync, readSync, closeSync, realpathSync } from "node:fs";
 import { execFile } from "node:child_process";
 import { extname, join, resolve as resolvePath, sep as pathSep } from "node:path";
+import { createChildProcessEnvironment } from "@bastani/atomic";
 import { activityMonitor } from "./activity.js";
 import type { ExtractedContent } from "./extract.js";
 import { flattenTruncatedString } from "./flat-string.js";
@@ -26,7 +27,7 @@ function cloneDir(config: GitHubCloneConfig, owner: string, repo: string, ref?:
 
 function execClone(args: string[], localPath: string, timeoutMs: number, signal?: AbortSignal): Promise<string | null> {
 	return new Promise((resolve) => {
-		const child = execFile(args[0], args.slice(1), { timeout: timeoutMs }, (err) => {
+		const child = execFile(args[0], args.slice(1), { timeout: timeoutMs, env: createChildProcessEnvironment() }, (err) => {
 			if (err) {
 				try {
 					rmSync(localPath, { recursive: true, force: true });

package/dist/builtin/web-access/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@bastani/web-access",
-  "version": "0.9.12",
+  "version": "0.9.13-alpha.2",
   "private": true,
   "description": "Atomic extension for web search, URL fetching, GitHub repo cloning, PDF/video extraction. Fork of: https://github.com/nicobailon/pi-web-access",
   "contributors": [
@@ -31,7 +31,7 @@
   },
   "peerDependencies": {
     "@bastani/atomic": "*",
-    "@earendil-works/pi-tui": "^0.83.0"
+    "@earendil-works/pi-tui": "^0.84.1"
   },
   "peerDependenciesMeta": {
     "@bastani/atomic": {
@@ -44,7 +44,7 @@
   "dependencies": {
     "@mozilla/readability": "^0.6.0",
     "linkedom": "^0.18.13",
-    "p-limit": "^7.3.0",
+    "p-limit": "^7.3.1",
     "turndown": "^7.2.0",
     "unpdf": "1.7.0"
   }

package/dist/builtin/web-access/subprocess.ts

@@ -1,3 +1,5 @@
+import { createChildProcessEnvironment } from "@bastani/atomic";
+
 export interface BunSubprocessOptions {
 	timeoutMs: number;
 	maxStdoutBytes: number;
@@ -60,7 +62,7 @@ export async function runBunSubprocess(
 	try {
 		proc = Bun.spawn([command, ...args], {
 			cwd: options.cwd,
-			env: options.env ? { ...process.env, ...options.env } : process.env,
+			env: createChildProcessEnvironment(options.env),
 			stdin: "ignore",
 			stdout: "pipe",
 			stderr: "pipe",

package/dist/builtin/web-access/summary-review.ts

@@ -1,5 +1,5 @@
 import type { ProviderHeaders } from "@earendil-works/pi-ai";
-import { complete, getModel, type Message, type Model } from "@earendil-works/pi-ai/compat";
+import { getModel, type Message, type Model } from "@earendil-works/pi-ai/compat";
 import type { ExtensionContext } from "@bastani/atomic";
 import type { QueryResultData } from "./storage.js";
 
@@ -256,7 +256,7 @@ export async function generateSummaryDraft(
 		timestamp: Date.now(),
 	};
 
-	const response = await complete(model, { messages: [userMessage] }, { apiKey, headers, signal });
+	const response = await ctx.modelRegistry.complete(model, { messages: [userMessage] }, { apiKey, headers, signal });
 	if (response.stopReason === "aborted") {
 		throw new Error("Aborted");
 	}

package/dist/builtin/web-access/web-search-features.ts

@@ -1,4 +1,4 @@
-import type { ExtensionAPI, ExtensionContext } from "@bastani/atomic";
+import { isStaleExtensionContextError, type ExtensionAPI, type ExtensionContext } from "@bastani/atomic";
 import { fetchAllContent } from "./extract.js";
 import { clearCloneCache } from "./github-extract.js";
 import { registerContentTools } from "./content-tools.js";
@@ -65,7 +65,7 @@ export function registerWebSearchFeatures(pi: ExtensionAPI, initConfig: WebSearc
 		const controller = new AbortController();
 		pendingFetches.set(fetchId, controller);
 		fetchAllContent(urls, controller.signal)
-			.then((fetched) => {
+			.then(async (fetched) => {
 				if (!runtimeState.sessionActive || !pendingFetches.has(fetchId)) return;
 				const data: StoredSearchData = {
 					id: fetchId,
@@ -74,33 +74,49 @@ export function registerWebSearchFeatures(pi: ExtensionAPI, initConfig: WebSearc
 					urls: stripThumbnails(fetched),
 				};
 				storeResult(fetchId, data);
-				pi.appendEntry("web-search-results", data);
-				const ok = fetched.filter(f => !f.error).length;
-				pi.sendMessage(
-					{
-						customType: "web-search-content-ready",
-						content: `Content fetched for ${ok}/${fetched.length} URLs [${fetchId}]. Full page content now available.`,
-						display: true,
-					},
-					{ triggerTurn: true },
-				);
+				try {
+					pi.appendEntry("web-search-results", data);
+					const ok = fetched.filter(f => !f.error).length;
+					// Keep the fetch pending until notification admission settles so session
+					// changes can still clear the entry and abort its controller while admission is in flight.
+					await pi.sendMessage(
+						{
+							customType: "web-search-content-ready",
+							content: `Content fetched for ${ok}/${fetched.length} URLs [${fetchId}]. Full page content now available.`,
+							display: true,
+						},
+						{ triggerTurn: true },
+					);
+				} catch (error) {
+					if (!isStaleExtensionContextError(error)) throw error;
+				}
 			})
-			.catch((err) => {
+			.catch(async (err) => {
+				if (isStaleExtensionContextError(err)) return;
 				if (!runtimeState.sessionActive || !pendingFetches.has(fetchId)) return;
 				const message = err instanceof Error ? err.message : String(err);
 				const isAbort = (err instanceof Error && err.name === "AbortError") || message.toLowerCase().includes("abort");
 				if (!isAbort) {
-					pi.sendMessage(
-						{
-							customType: "web-search-error",
-							content: `Content fetch failed [${fetchId}]: ${message}`,
-							display: true,
-						},
-						{ triggerTurn: false },
-					);
+					try {
+						await pi.sendMessage(
+							{
+								customType: "web-search-error",
+								content: `Content fetch failed [${fetchId}]: ${message}`,
+								display: true,
+							},
+							{ triggerTurn: false },
+						);
+					} catch (error) {
+						if (!isStaleExtensionContextError(error)) throw error;
+					}
 				}
 			})
-			.finally(() => { pendingFetches.delete(fetchId); });
+			.finally(() => { pendingFetches.delete(fetchId); })
+			.catch((error) => {
+				if (!isStaleExtensionContextError(error)) {
+					console.error("Web search background fetch notification failed:", error);
+				}
+			});
 		return fetchId;
 	}
 

package/dist/builtin/web-access/web-search-summary.ts

@@ -1,4 +1,4 @@
-import { complete, getModel, type Model } from "@earendil-works/pi-ai/compat";
+import { getModel, type Model } from "@earendil-works/pi-ai/compat";
 import type { ProviderHeaders } from "@earendil-works/pi-ai";
 import type { QueryResultData } from "./storage.js";
 import {
@@ -39,7 +39,7 @@ export async function rewriteSearchQuery(
 		{ provider: "google", id: "gemini-2.5-flash" },
 		{ provider: "openai", id: "gpt-4.1-mini" },
 	]);
-	const response = await complete(
+	const response = await ctx.modelRegistry.complete(
 		model,
 		{
 			messages: [{

package/dist/builtin/workflows/CHANGELOG.md

@@ -6,6 +6,76 @@ The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.1.0/),
 
 ## [Unreleased]
 
+## [0.9.13-alpha.2] - 2026-08-12
+
+### Breaking Changes
+
+- Custom workflow UI components must return `true` when they consume input and `false` when they do not. In fullscreen mode, an unhandled result lets the transcript process matching viewport navigation keys instead of swallowing them.
+- Added `failed` to the `ctx.exit()` status union. Author-initiated failed exits persist their reason and partial declared outputs, return the discriminated `{ exited: true, status: "failed" }` child result instead of throwing, and default to non-resumable unless `resumable: true` is set. Downstream exhaustive switches over `WorkflowExitStatus` must handle the new failed-exit case ([#2242](https://github.com/bastani-inc/atomic/issues/2242)).
+
+### Changed
+
+- Updated the Ralph research stage model configuration. The primary model moves from `openai-codex/gpt-5.6-luna:max` to `anthropic/claude-opus-5:high`, and the fallback chain is rebuilt around high/xhigh thinking levels: GPT-5.6 Sol at `xhigh` replaces the Luna variants, Claude Fable 5 and Claude Opus 4.8 step up from `low`/`medium` to `high`, GPT-5.5 and GLM-5.2 step up to `xhigh`, and Kimi K3 (`kimi-coding`, `moonshotai`, `moonshotai-cn`, and OpenRouter) plus `openrouter/sakana/fugu-ultra:high` join the chain.
+- Extended workflow-authoring guidance with a stacked implementation-slices starter pattern: authors target roughly 100–500 changed lines between verification points by default, route each slice through a child lifecycle, explicitly create or check out each named branch before its child because `base_branch` and `git_worktree_dir` alone do not create feature branches, stack each verified branch and worktree, and stop at the first failed gate while preserving earlier verification ([#2262](https://github.com/bastani-inc/atomic/issues/2262)).
+- Updated authored-workflow model guidance to treat benchmark thinking levels as measurement settings, assign effort by stage role and failure cost, require a compact `Stage | Model | Thinking | Role` declaration before launch, reserve `max` for high-cost-of-error roles or explicit user requests, and apply catalog-aware role policy to primary and fallback attempts ([#2203](https://github.com/bastani-inc/atomic/issues/2203)).
+- Added bounded constructive-quorum deliberation to Goal and Ralph reviewer prompts: reviewers form independent preliminary verdicts, exchange evidence once over Intercom, and record evidence-backed verdict changes before the existing reducer counts final votes ([#2237](https://github.com/bastani-inc/atomic/issues/2237)).
+
+- Rebuilt the workflow graph overlay on pi-tui 0.84.1's `VStack` and `ScrollView` layout primitives. The graph body now adapts to terminal height and resize, supports vertical wheel scrolling and a draggable scrollbar, and keeps wide-graph horizontal panning and existing prompt/navigation key handling ([#2223](https://github.com/bastani-inc/atomic/issues/2223)).
+- Retired the legacy terminal-row callback thread from graph and stage-chat views. The two fullscreen-frame consumers still read `terminal.rows` directly: GraphView uses it for the overlay frame, and StageChatView uses it for the chat line count, falling back to the unhosted `VIEW_LINE_COUNT = 32` frame when no host is present. `PiCustomComponent.render` supplies only a width, so the #2223 viewport-row criterion remains partial at those call paths ([#2223](https://github.com/bastani-inc/atomic/issues/2223)).
+
+- Workflow command, worktree Git/setup-hook, and Playwright CLI subprocesses now receive `AI_AGENT=atomic` for generic child-process attribution without mutating caller-supplied environment objects.
+
+### Fixed
+- Fixed completed `ctx.exit()` runs losing the reserved returned-status convention, legacy child replay records losing their exit discriminator, and failed author-exit resume surfaces falling back to snapshot inspection instead of durable retry.
+- Fixed workflow graph node cards appending the Codex `fast` marker to dependency metadata; cards now show only topology details such as `root`, `1 dep`, or `<n> deps` while fast-mode state remains available to the stage and its model label ([#2337](https://github.com/bastani-inc/atomic/issues/2337)).
+
+- Fixed the workflow graph overlay statusline echoing the `i-have-adhd` extension's `ADHD Mode` badge. The badge remains in the main-chat footer while other extension statuses continue to render in the overlay ([#2341](https://github.com/bastani-inc/atomic/issues/2341)).
+
+- Fixed fullscreen workflow graph and stage-chat mouse routing. Focused workflow overlays now receive SGR/X10 wheel and click press/release sequences through the host's bounded custom-input reply channel before pi-tui's alternate-screen viewport, restoring graph scrolling and click-to-attach without a workflow-owned TTY-tracking seam; the fullscreen route also mirrors consumed left-button events into pi-tui's application-owned selection path ([#2303](https://github.com/bastani-inc/atomic/issues/2303)).
+- Fixed graph overlay rendering to keep its live vertical position in pi-tui's `ScrollView` and preserve OSC-8 hyperlink terminators when normalizing layout rows.
+
+- Fixed switcher wheel input scrolling an obscured graph and the scrollbar covering the final graph column; overflow now reserves its scrollbar column and the switcher owns wheel selection ([#2223](https://github.com/bastani-inc/atomic/issues/2223)).
+- Restored natural graph geometry for unhosted overlays by removing the fixed-height fallback and kept pi-tui's layout metadata aligned with sparse, deeply scrolled content, so large graphs track their natural geometry without corrupting OSC-8 rows ([#2223](https://github.com/bastani-inc/atomic/issues/2223)).
+
+- Fixed the inline workflow form editor not inheriting the host's `autocompleteMaxVisible` setting when Atomic installs it as a custom editor.
+- Fixed late workflow messages and MCP scope cleanup that finish after an extension reload. Stale event-bus calls now stop without retrying through stale APIs, while active routes keep their normal fallback behavior.
+- Fixed workflow stale-context guards by consuming the host's exported predicate instead of copying its error-message marker.
+- Fixed late-message routing to reject stale drops consistently for Intercom and generic batches, so a resolved route means delivery was accepted instead of silently treating a dropped message as routed; non-stale failures still propagate.
+
+- Fixed the `DISPATCHED` confirmation panel being torn apart by any workflow input containing a newline. Input values were truncated by visible width, which a control character does not have, so an embedded newline survived truncation and emitted extra physical lines that no border ever wrapped — the box was destroyed from that row down. Multi-line inputs are ordinary, so this fired for most real launches: a `prompt` or an `acceptance_criteria` block reliably reproduced it. String values and the object/array JSON projection now collapse control characters, `U+2028`, and `U+2029` to single spaces before truncation, so a run card stays one row per value. `U+2028`/`U+2029` are included because `JSON.stringify` does not escape them and some terminals still break lines on them. As a side effect, escape sequences in an input value can no longer inject ANSI styling into the chat surface.
+- Fixed workflow stages being unable to use the `subagent` tool at all. The stage policy set `managementActions: "full"` and `fanoutAuthorized: false` together, which contradict: stages could neither delegate nor, because of the companion subagents defect, run read-only management such as `subagent list`. Workflow stages are top-level sessions rather than subagent children, so the policy now sets `fanoutAuthorized: true`, restoring the delegation the workflow docs already describe. Nesting remains bounded by the typed in-process depth policy and Rust admission’s five-level ceiling ([#2220](https://github.com/bastani-inc/atomic/pull/2220), regression from [#2205](https://github.com/bastani-inc/atomic/pull/2205)).
+- Fixed nested subagents in workflow-stage sessions losing the bundled `subagent` extension. In-process child resource loading now carries the bundled package roots and disables only the workflow extension's repeated stage lifecycle, so a stage can delegate and its nested child receives the bundled tools instead of only the base built-in ones ([#2220](https://github.com/bastani-inc/atomic/pull/2220), regression from [#2205](https://github.com/bastani-inc/atomic/pull/2205)).
+- Fixed a stage chat opened on a stage that was already mid-turn showing an empty transcript. The stage's assistant message joins the handle's message list only when its turn ends, and streaming updates now carry a delta rather than the message so far, so a chat mounted after the stream started had nothing to draw and nothing that would restate it — a stage held mid-turn stayed blank for the whole turn. The chat now seeds itself from the stage session's in-flight assistant message when it mounts, and the deltas that arrive after that continue the same message.
+
+### Removed
+
+- Removed workflow-owned mouse-scroll reporting and attached-stage copy mode. Fullscreen pi-tui now keeps application-owned drag and multi-click selection available over workflow overlays, and `ctrl+t` is left to the host keybindings. Copy still uses OSC 52; terminals that refuse OSC 52 writes require the modifier-drag bypass (Shift/Option, as provided by the terminal).
+
+## [0.9.13-alpha.1] - 2026-08-05
+
+### Added
+
+- Added four user-initiated workflow lifecycle notices — `started`, `paused`, `quit`, and `resumed` — so the main chat learns when a top-level background run begins, stops, or picks back up. Each travels the same steer delivery, capped-backoff retry, and notice-card path the failure notice already uses, so an agent waiting on a background run no longer mistakes a deliberate stop for work still in progress. `WORKFLOW STARTED` (`▶`), `WORKFLOW PAUSED` (`⏸`, warning tone), `WORKFLOW QUIT` (`⏹`, warning tone, also reporting whether the run stayed resumable), and `WORKFLOW RESUMED` (`▶`) render as cards; the paused and quit text states that the stop was deliberate and user-requested and tells the model not to resume the run or take over the work unless asked, hinting `/workflow resume <run-id>`, while the resumed text does not, because the run is progressing again ([#2177](https://github.com/bastani-inc/atomic/issues/2177)).
+- Only user actions raise those notices. `/workflow <name>`, `/workflow pause`, `/workflow quit`, and `/workflow resume` notify; the matching `workflow({ action: "run" | "pause" | "quit" | "resume" })` tool calls stay silent because the tool result already reports them, and `/workflow interrupt` raises nothing. Engine-internal transitions stay silent as well: a notice exists only where a control path recorded an actor, which is what keeps answering a human-in-the-loop prompt, per-stage control, and the resume-acknowledgement pass out of the chat ([#2177](https://github.com/bastani-inc/atomic/issues/2177)).
+- Lifecycle notices now carry two attributions. *Origin* — who launched the run — renders on every kind as "which you started" or "which the user started", is recorded once at dispatch, survives session restore and durable resume, and is inherited by a continuation from the run it continues rather than recomputed. *Actor* — who performed this one event — renders as "The user paused". A run with no recorded origin, including a legacy or restored snapshot, omits the clause entirely instead of guessing ([#2177](https://github.com/bastani-inc/atomic/issues/2177)).
+- One request now produces exactly one notice. A whole-run pause or resume reports at run scope; a stage-scoped pause or resume that leaves siblings paused reports at stage scope, and one that stops or restarts the whole run reports the run instead, never both. A quit reports the quit alone rather than the pause it publishes on the way. Because control actions are reversible, they are deduplicated by run id together with the occurrence timestamp, so pause → resume → pause → resume notifies four times while repeated snapshot invalidations at one unchanged state notify once. Resuming reports a resume and never a start, whoever requested it: resuming a failed or blocked run launches a continuation under a fresh id whose notice names both run ids, while resuming a quit run reuses the original workflow id so durable checkpoints replay and its notice names the one id. A run already started, paused, or quit when notifications install (restore, replay, `/reload`, session-preserving reinstall) stays silent, and nested child runs never notify at top level. The default `workflowNotifications.notifyOn` becomes `["started", "completed", "failed", "blocked", "awaiting_input", "paused", "quit", "resumed"]`; a config that pins `notifyOn` explicitly keeps exactly the kinds it lists ([#2177](https://github.com/bastani-inc/atomic/issues/2177)).
+
+### Changed
+
+- Workflow stage resource loading now carries typed subagent policy in session options and no longer mutates process-global child environment flags during reload. Concurrent stage creation preserves tool allowlists and orchestration limits without the process-era isolation race ([#2188](https://github.com/bastani-inc/atomic/issues/2188)).
+
+### Fixed
+
+- Fixed thrown transient stage failures — rate limits, provider outages, network timeouts, transport errors — to retry the current model with bounded exponential backoff from `settings.retry` before advancing through `fallbackModels`; disabling retry keeps immediate candidate advancement. Failures another request to the same model cannot repair — a rejected credential, an unavailable model, an incompatible request — skip the same-model retry and advance to the next candidate immediately, for prompt delivery and session creation alike ([#2170](https://github.com/bastani-inc/atomic/issues/2170)).
+- Fixed a same-candidate stage retry losing the stage prompt. The retry resumed the existing session turn with a transcript that no longer ended in the prompt it was resuming, which the agent rejects; the prompt is now retained for that path and still re-sent on the re-prompt path ([#2170](https://github.com/bastani-inc/atomic/issues/2170)).
+- Fixed a same-candidate stage retry stopping outright when the failed attempt had already streamed a completed assistant message. Resuming the existing turn is now attempted only when the transcript, as converted for the provider, still ends in a message the agent can continue from; otherwise the retry re-sends the prompt, which recovers instead of failing. Messages that convert away — a note excluded from context, an empty branch summary — no longer look like a valid tail ([#2170](https://github.com/bastani-inc/atomic/issues/2170)).
+- Fixed an eagerly created stage session never reaching its `fallbackModels`. A session created before any prompt — by an eager stage call, a control attach, or `ctx.__ensureSession()` — retried only its first candidate and then failed the stage. It now walks the configured chain, and a failure that exhausts the chain is no longer cached and replayed to every later caller ([#2170](https://github.com/bastani-inc/atomic/issues/2170)).
+- Fixed a paused eager session creation dropping its replacement objective when the stage had explicit model candidates. The resumed text is now authoritative for the next prompt and is sent exactly once ([#2170](https://github.com/bastani-inc/atomic/issues/2170)).
+- Fixed a context overflow that compaction had already failed to resolve being retried on the same candidate for the whole `settings.retry` budget. Re-sending an identical request cannot fit a context compaction could not shrink, so the chain now advances immediately while still treating the failure as fallbackable ([#2170](https://github.com/bastani-inc/atomic/issues/2170)).
+- Fixed a stage running two candidate walks at once. Advancing a candidate released the shared creation gate while the walk was still creating the next session, so a concurrent `ctx.__ensureSession()` or a first `ctx.prompt()` could start a second walk, duplicate provider work, and leave the losing session live and undisposed. Concurrent callers now join the creation already in flight ([#2170](https://github.com/bastani-inc/atomic/issues/2170)).
+- Fixed a controlled pause that both started and finished while a stage session was still being created losing its replacement objective, so the stale objective was prompted instead ([#2170](https://github.com/bastani-inc/atomic/issues/2170)).
+- Fixed a controlled pause during a same-candidate continuation aborting the stage unrecoverably. The abort a pause causes is no longer classified as a terminal model failure: the pause settles, and resuming with a replacement objective re-prompts on the same candidate instead of spending a fallback ([#2170](https://github.com/bastani-inc/atomic/issues/2170)).
+
 ## [0.9.12] - 2026-08-04
 
 Cumulative release of the `0.9.12-alpha.1` prerelease. The summary below covers the user-visible outcome of that work; the per-change detail remains in the prerelease section below.