@bastani/atomic 0.9.3-alpha.6 → 0.9.4-alpha.1

package/CHANGELOG.md

@@ -2,6 +2,46 @@
 
 ## [Unreleased]
 
+## [0.9.4-alpha.1] - 2026-06-29
+
+### Changed
+
+- Restored upstream Pi's provider retry default by leaving `retry.provider.maxRetries` unset/zero unless users configure it explicitly, so Atomic's agent-level retry observes provider transport failures directly instead of the SDK retrying them first.
+
+### Fixed
+
+- Aligned Atomic's global HTTP dispatcher with upstream Pi by removing the fixed 10-second connect-phase timeout and installing undici's fetch implementation alongside the configured proxy dispatcher, avoiding spurious provider `Connection error` failures behind policy/proxy egress layers.
+- Made headless JSON print mode return a nonzero exit code when the final assistant turn ends with `stopReason: "error"` or `"aborted"`, so Pier and other harnesses can mark provider failures as agent errors.
+- Declared `lru-cache` as a direct runtime dependency and added publish-time `npm-shrinkwrap.json` generation, matching upstream Pi's deterministic npm installs and fixing strict pnpm global installs that could not resolve Atomic's hashline and fetch-tool cache imports.
+- Updated user-facing workflow documentation for the blocked lifecycle notice state, the default `workflowNotifications.notifyOn` set, and the reserved top-level workflow output `status` convention for returned `failed`/`blocked` run statuses.
+
+## [0.9.3] - 2026-06-29
+
+### Breaking Changes
+
+- Replaced the legacy exact-replacement `edit` tool API with the hashline-only `input` script schema; callers must use snapshot tags from `read`/`search`/`write`/`edit` instead of `path` + `edits[]` or `oldText`/`newText`.
+- Tightened the model-facing `read`, `find`, and `search` schemas to the new builtin contracts (`read` path selectors, required `find.paths`, and the narrowed `search` option set).
+
+### Added
+
+- Added first-run onboarding that routes pasted tickets, specs, and tasks into normal Atomic sessions with scope-estimation guidance and workflow handoff next steps.
+- Added internal workflow-stage session marking so workflow-created sessions stay out of ordinary `/resume`, `atomic -r`, and `--continue` history while remaining accessible through workflow resume/status surfaces.
+- Added first-class `find`/`search` builtins, hashline snapshot anchors, hashline edit scripts with stale-tag safety, a disabled-by-default Bash Interceptor toggle, Rust-backed `pty:true` bash execution, and native glob/grep/search bindings for oh-my-pi parity.
+
+### Changed
+
+- Synced bundled upstream Pi runtime packages to `^0.80.2` and routed legacy pi-ai imports through the temporary `/compat` entrypoint, including virtual-module and Jiti aliases for first-party and user-installed extensions.
+- Enabled provider/SDK retries by default, raised the standard HTTP idle timeout to 10 minutes, and added a 10-second connect-phase timeout so transient socket drops retry while unreachable hosts fail fast.
+- Accounted for image content blocks in context-window estimates and compaction budgets, raised delegated workflow/subagent nesting to the shared five-level maximum, and exported the canonical schema-aware flattened-argument helper used by host and MCP tool execution.
+- Updated workflow and user-facing documentation so structured, validation-heavy, implementation, debugging, migration, and loop-shaped requests are routed to workflows by default.
+
+### Fixed
+
+- Fixed custom tool renderer cleanup, persisted-context replay after context compaction, and multiple first-run onboarding edge cases around placeholders, file references, isolated config discovery, saved tasks, `/import`, `/model`, and `/new`.
+- Completed the oh-my-pi builtin parity pass across `read`, `write`, `find`, `search`, `edit`, `bash`, archive, document, URL, internal-resource, conflict, and SQLite selectors, including pagination, truncation metadata, native cache invalidation, copied-hashline stripping, async/PTY behavior, and cross-platform path handling.
+- Hardened generated-file, URL, `local://`, SQLite, archive, native PTY/search, and bash-interceptor paths against traversal, SSRF, unsafe raw SQL/archive cases, stale cache writes, native panics, leaked async output, and provider-hostile argument shapes.
+- Fixed compiled release binary packaging and cross-platform package tests by externalizing `mupdf`, preparing native bindings and fixtures in CI, running the coding-agent Vitest suite under Bun, and hardening Windows path, color, and process-spawn coverage.
+
 ## [0.9.3-alpha.6] - 2026-06-29
 
 ### Changed

package/dist/builtin/cursor/CHANGELOG.md

@@ -2,6 +2,36 @@
 
 ## [Unreleased]
 
+## [0.9.4-alpha.1] - 2026-06-29
+
+### Changed
+
+- Published a synchronized Atomic 0.9.4-alpha.1 prerelease for the Cursor provider package; no functional Cursor provider changes were made after 0.9.3.
+
+## [0.9.3] - 2026-06-29
+
+### Added
+
+- Added scoped Cursor image-input support for known multimodal Claude, Composer, Gemini, GPT, Kimi, and Grok 4.3 model families, including selected-image request serialization and mixed text/image MCP tool-result serialization.
+- Added `cursor/grok-4.3` to the estimated fallback catalog.
+
+### Changed
+
+- Aligned the Cursor provider dependency on upstream pi-ai `^0.80.2` and retargeted legacy provider/model imports to the `@earendil-works/pi-ai/compat` entrypoint.
+- Resolved Cursor model context windows and max output tokens from Atomic's bundled pi-ai model catalog while preserving positive live limits, ignoring bogus non-positive values, and keeping conservative estimates only for Cursor-only models.
+- Marked Cursor OAuth as **Cursor (Experimental)** in the `/login` picker.
+
+### Fixed
+
+- Fixed effort-variant-only Cursor models failing no-thinking requests by recording and sending a concrete default variant instead of an unsendable synthesized base id.
+- Preserved explicit `1M` Cursor context floors across fallback catalog matches and made the pi-ai catalog a runtime dependency so limit fallback remains available at provider startup.
+- Exposed `xhigh` only for Cursor models whose live or estimated catalog includes a real `xhigh`/`max` variant, with saved `xhigh` selections falling back to the nearest sendable variant.
+- Rejected empty or malformed base64 image payloads during Cursor selected-image and MCP image serialization while accepting valid MIME-wrapped base64 with whitespace.
+
+### Removed
+
+- Removed outdated Cursor Grok 4.20 fallback entries and stopped advertising unsupported Grok-family Cursor IDs as image-capable.
+
 ## [0.9.3-alpha.6] - 2026-06-29
 
 ### Changed

package/dist/builtin/cursor/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@bastani/cursor",
-  "version": "0.9.3-alpha.6",
+  "version": "0.9.4-alpha.1",
   "private": true,
   "description": "Experimental first-party Atomic extension for Cursor OAuth, model discovery, and streaming provider registration.",
   "contributors": [
@@ -40,7 +40,7 @@
     }
   },
   "dependencies": {
-    "@bastani/atomic-natives": "0.9.3-alpha.6",
+    "@bastani/atomic-natives": "0.9.4-alpha.1",
     "@bufbuild/protobuf": "^2.0.0",
     "@earendil-works/pi-ai": "^0.80.2"
   }

package/dist/builtin/intercom/CHANGELOG.md

@@ -4,6 +4,18 @@ All notable changes to the `pi-intercom` extension will be documented in this fi
 
 ## [Unreleased]
 
+## [0.9.4-alpha.1] - 2026-06-29
+
+### Changed
+
+- Published a synchronized Atomic 0.9.4-alpha.1 prerelease for the intercom extension; no intercom extension changes were made after 0.9.3.
+
+## [0.9.3] - 2026-06-29
+
+### Changed
+
+- Published the stable Atomic 0.9.3 release for the intercom extension with its upstream pi TUI peer dependency aligned to `^0.80.2`; no intercom extension source changes were needed after 0.9.2.
+
 ## [0.9.3-alpha.6] - 2026-06-29
 
 ### Changed

package/dist/builtin/intercom/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@bastani/intercom",
-  "version": "0.9.3-alpha.6",
+  "version": "0.9.4-alpha.1",
   "private": true,
   "description": "Atomic extension providing a private coordination channel between parent and child agent sessions. Fork of: https://github.com/nicobailon/pi-intercom",
   "contributors": [

package/dist/builtin/mcp/CHANGELOG.md

@@ -7,6 +7,22 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ## [Unreleased]
 
+## [0.9.4-alpha.1] - 2026-06-29
+
+### Changed
+
+- Published a synchronized Atomic 0.9.4-alpha.1 prerelease for the MCP extension; no MCP extension changes were made after 0.9.3.
+
+## [0.9.3] - 2026-06-29
+
+### Changed
+
+- Aligned the MCP extension peer dependencies with upstream pi `^0.80.2` and retargeted legacy provider/model imports to `@earendil-works/pi-ai/compat`, preserving MCP sampling behavior under the updated pi-ai package layout.
+
+### Fixed
+
+- Made `unflattenToolArguments` schema-aware so literal dotted top-level argument keys such as `filter.name` are preserved unless the tool input schema proves the key represents a nested path, while bracket-indexed array/object keys are still reconstructed for direct-tool and proxy/gateway calls.
+
 ## [0.9.3-alpha.6] - 2026-06-29
 
 ### Changed

package/dist/builtin/mcp/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@bastani/mcp",
-  "version": "0.9.3-alpha.6",
+  "version": "0.9.4-alpha.1",
   "private": true,
   "description": "Atomic extension that adapts MCP (Model Context Protocol) servers into the coding agent. Fork of: https://github.com/nicobailon/pi-mcp-adapter",
   "contributors": [

package/dist/builtin/subagents/CHANGELOG.md

@@ -2,6 +2,29 @@
 
 ## [Unreleased]
 
+## [0.9.4-alpha.1] - 2026-06-29
+
+### Changed
+
+- Published a synchronized Atomic 0.9.4-alpha.1 prerelease for the subagents extension; no subagents extension changes were made after 0.9.3.
+
+## [0.9.3] - 2026-06-29
+
+### Added
+
+- Added OpenRouter fallback coverage to bundled subagent definitions so delegated codebase analysis, research, debugging, simplification, and worker sessions can recover through OpenRouter-hosted frontier, GLM, and Gemini providers when native providers are unavailable.
+
+### Changed
+
+- Aligned the subagents extension peer dependencies with upstream pi `^0.80.2` runtime packages and retargeted legacy provider/model imports to `@earendil-works/pi-ai/compat`.
+- Raised the default and hard maximum subagent nesting budget to five delegated levels, extending nested-run observability to the same depth.
+
+### Fixed
+
+- Prevented async and foreground subagent status widgets from flickering or unmounting during host UI refreshes, tall-panel redraws, and reset-and-hydrate cycles with active background runs.
+- Fixed live subagent result animation cleanup so terminal workflow rows evict animation registry entries instead of leaving stale intervals.
+- Synced upstream subagent hardening for compact tool-call summaries, nested fanout call/result history, duplicate dispatch rejection, provider-safe chain schemas, failed foreground diagnostics, builtin `search` MCP allowlist filtering, and updated builtin `find`/`search` tool guidance.
+
 ## [0.9.3-alpha.6] - 2026-06-29
 
 ### Added

package/dist/builtin/subagents/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@bastani/subagents",
-  "version": "0.9.3-alpha.6",
+  "version": "0.9.4-alpha.1",
   "private": true,
   "description": "Atomic extension for delegating tasks to subagents with chains, parallel execution, and TUI clarification. Fork of: https://github.com/nicobailon/pi-subagents",
   "contributors": [

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

@@ -4,6 +4,18 @@ All notable changes to this project will be documented in this file.
 
 ## [Unreleased]
 
+## [0.9.4-alpha.1] - 2026-06-29
+
+### Changed
+
+- Published a synchronized Atomic 0.9.4-alpha.1 prerelease for the web-access extension; no web-access extension changes were made after 0.9.3.
+
+## [0.9.3] - 2026-06-29
+
+### Changed
+
+- Published the stable Atomic 0.9.3 release for the web-access extension with its upstream pi TUI peer dependency aligned to `^0.80.2` and legacy pi-ai summary/model helper imports retargeted to the `/compat` entrypoint; no web-access extension source changes were needed after 0.9.2.
+
 ## [0.9.3-alpha.6] - 2026-06-29
 
 ### Changed

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

@@ -1,6 +1,6 @@
 {
   "name": "@bastani/web-access",
-  "version": "0.9.3-alpha.6",
+  "version": "0.9.4-alpha.1",
   "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": [

package/dist/builtin/workflows/CHANGELOG.md

@@ -6,6 +6,38 @@ The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.1.0/),
 
 ## [Unreleased]
 
+## [0.9.4-alpha.1] - 2026-06-29
+
+### Fixed
+
+- Fixed workflow lifecycle notices for structured workflow outputs whose returned `status` is `failed` or `blocked`: the runtime now records those runs as failed/blocked instead of successful completions, preserves their returned output, carries blocked summaries into lifecycle notice reasons, marks returned failures non-resumable with explicit terminal failure metadata, and emits failure/blocked lifecycle cards rather than success/checkmark notices.
+
+## [0.9.3] - 2026-06-29
+
+### Breaking Changes
+
+- Enabled workflow durability by default with a zero-infrastructure per-workflow file backend under `~/.atomic/workflow-durable`, replacing the previous process-local in-memory default and removing the old `ATOMIC_WORKFLOW_DURABLE_DIR` opt-in path. Explicit in-memory durability remains available through custom/test backends or `ATOMIC_WORKFLOW_DURABLE=0`/`false`/`off`/`memory`.
+
+### Added
+
+- Added `StageContext.sendUserMessage(content, { deliverAs? })` so workflow authors can inject follow-on user turns into live stage sessions after `prompt()` resolves while preserving MCP scope, abort handling, and structured-output guards.
+- Added cross-session durable workflow resumability with cached `ctx.tool`, `ctx.ui`, `ctx.stage`/`ctx.task`/`ctx.chain`/`ctx.parallel`, and child-workflow checkpoints, plus `/workflow resume` discovery of durable workflow history from session metadata and the durable backend.
+- Added optional DBOS-backed durability through the lazily loaded `@dbos-inc/dbos-sdk`, including DBOS read-side hydration for resumable workflows when `DBOS_SYSTEM_DATABASE_URL` is set.
+- Added OpenRouter fallback coverage to builtin `goal`, `ralph`, `deep-research-codebase`, and `open-claude-design` workflow model configs.
+
+### Changed
+
+- Marked workflow stage sessions as internal so they are excluded from normal Atomic resume history while remaining inspectable and resumable through workflow-specific commands and tools.
+- Revised builtin `goal`, `ralph`, and `open-claude-design` prompts to emphasize full objective completion, stable non-ordinal artifacts, a shared five-level subagent nesting budget, and model-facing workflow guidance that routes non-trivial structured work to workflows by default.
+- Aligned the workflows extension peer dependency with upstream pi TUI `^0.80.2`.
+
+### Fixed
+
+- Fixed workflow graph and attached-stage chat interactions, including direct mouse-click activation, mouse/trackpad wheel capture, copy-mode toggling, stale `Working...` spinners, terminal-row animation cleanup, and post-terminal subagent progress handling.
+- Hardened durable resume, replay, checkpoint, and listing behavior across file and DBOS backends, including pending prompts, terminal/non-resumable states, stale cache entries, repeated quit/resume cycles, child workflow scope isolation, checkpoint identity hashing, retries, cancellation, and durable write failures.
+- Fixed `/workflow resume` picker behavior for live, paused, failed, and cross-session durable workflows so it matches the standard Atomic resume selector and handles empty, headless, mixed live/durable, and cache-only states consistently.
+- Fixed builtin workflow continuity and configuration edge cases, including `open-claude-design` generator/feedback forking, isolated workflow config discovery, durable schema-backed stage replay, parallel fail-fast finalization, and custom-backend propagation into child workflows.
+
 ## [0.9.3-alpha.6] - 2026-06-29
 
 ### Added

package/dist/builtin/workflows/README.md

@@ -31,18 +31,18 @@ Adding workflow files under `.atomic/workflows/` (project scope) or `~/.atomic/a
 
 ### Workflow lifecycle notifications
 
-Workflow lifecycle notices are enabled by default. They send steer prompts into the main chat/model context when a run completes or fails. Awaiting-input prompts are tracked for dedupe/restore, but they do not wake the main chat agent. Configure lifecycle tracking in the same extension config file:
+Workflow lifecycle notices are enabled by default. They send steer prompts into the main chat/model context when a run completes, fails, or ends blocked. Awaiting-input prompts are tracked for dedupe/restore, but they do not wake the main chat agent. Configure lifecycle tracking in the same extension config file:
 
 ```json
 {
   "workflowNotifications": {
     "enabled": true,
-    "notifyOn": ["completed", "failed", "awaiting_input"]
+    "notifyOn": ["completed", "failed", "blocked", "awaiting_input"]
   }
 }
 ```
 
-Set `enabled` to `false` to disable all lifecycle notices, or narrow `notifyOn` to a non-empty list of selected events. Completion and failure lifecycle notices are emitted for top-level workflow runs, use steer delivery, and wake an idle model so the lifecycle update enters the model context when it happens. Nested child workflow completion/failure is reflected inside the expanded parent graph instead of producing separate top-level completion cards. Awaiting-input states are tracked for dedupe/restore, but workflows do not enqueue main-chat `/workflow connect` cards for them; prompt state remains visible through workflow status/connect surfaces, avoiding stale actionable cards if a prompt resolves while the main chat is streaming.
+Set `enabled` to `false` to disable all lifecycle notices, or narrow `notifyOn` to a non-empty list of selected events. Completion, failure, and blocked lifecycle notices are emitted for top-level workflow runs, use steer delivery, and wake an idle model so the lifecycle update enters the model context when it happens. Nested child workflow completion/failure is reflected inside the expanded parent graph instead of producing separate top-level completion cards. Awaiting-input states are tracked for dedupe/restore, but workflows do not enqueue main-chat `/workflow connect` cards for them; prompt state remains visible through workflow status/connect surfaces, avoiding stale actionable cards if a prompt resolves while the main chat is streaming.
 
 When a stage human-in-the-loop prompt is answered from the workflow TUI/stage chat, workflows also emits a separate display-only `workflows:hil-answer-notice` custom message. It records the answer for user-visible audit, but it does not wake the main agent, enter LLM context, or authorize answering later workflow prompts. Answers sent by the main-chat `workflow` tool do not emit this notice because the tool result already tells the main agent what happened.
 
@@ -736,7 +736,7 @@ Config-based discovery (`~/.atomic/agent/extensions/workflow/config.json` or `.a
   },
   "workflowNotifications": {
     "enabled": true,
-    "notifyOn": ["completed", "failed", "awaiting_input"]
+    "notifyOn": ["completed", "failed", "blocked", "awaiting_input"]
   }
 }
 ```

package/dist/builtin/workflows/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@bastani/workflows",
-  "version": "0.9.3-alpha.6",
+  "version": "0.9.4-alpha.1",
   "private": true,
   "description": "Atomic extension for multi-stage workflow authoring and execution.",
   "contributors": [

package/dist/builtin/workflows/src/engine/run-durable-finalize.ts

@@ -33,11 +33,13 @@ export async function finalizeDurableTerminalStatus(input: DurableTerminalFinali
   if (!input.isRoot) return;
   const status = input.runSnapshot.status;
   const isExitTerminal = input.runSnapshot.exited === true && status !== "running";
-  if (status !== "failed" && status !== "killed" && !isExitTerminal) return;
+  const isReturnedBlockedTerminal = status === "blocked" && input.runSnapshot.endedAt !== undefined;
+  if (status !== "failed" && status !== "killed" && !isExitTerminal && !isReturnedBlockedTerminal) return;
 
   const durableStatus = toDurableStatus(status);
   if (durableStatus !== undefined) {
-    input.durableBackend.setWorkflowStatus(input.runId, durableStatus, undefined, input.runSnapshot.resumable);
+    const resumable = status === "blocked" ? false : input.runSnapshot.resumable;
+    input.durableBackend.setWorkflowStatus(input.runId, durableStatus, undefined, resumable);
   }
   try {
     await input.durableBackend.flush?.();

package/dist/builtin/workflows/src/engine/run-returned-status.ts

@@ -0,0 +1,41 @@
+import type { RunEndMetadata } from "../shared/store-public-types.js";
+import type { WorkflowOutputValues } from "../shared/types.js";
+
+export interface ReturnedRunStatus {
+  readonly status: "completed" | "failed" | "blocked";
+  readonly error?: string;
+  readonly metadata?: RunEndMetadata;
+}
+
+export function classifyReturnedRunStatus(result: WorkflowOutputValues | undefined): ReturnedRunStatus {
+  const returnedStatus = result?.["status"];
+  if (returnedStatus !== "failed" && returnedStatus !== "blocked") {
+    return { status: "completed" };
+  }
+
+  const summary = result?.["summary"];
+  const error = typeof summary === "string" && summary.trim().length > 0
+    ? summary.trim()
+    : `Workflow returned status ${JSON.stringify(returnedStatus)}.`;
+  return {
+    status: returnedStatus,
+    error,
+    metadata: returnedStatus === "failed"
+      ? returnedFailureMetadata(error)
+      : returnedBlockedMetadata(),
+  };
+}
+
+function returnedFailureMetadata(error: string): RunEndMetadata {
+  return {
+    failureKind: "unknown",
+    failureRecoverability: "non_recoverable",
+    failureDisposition: "terminal_failed",
+    failureMessage: error,
+    resumable: false,
+  };
+}
+
+function returnedBlockedMetadata(): RunEndMetadata {
+  return { resumable: false };
+}

package/dist/builtin/workflows/src/engine/run.ts

@@ -41,6 +41,7 @@ import {
 import { assertWorkflowRunOutputs, normalizeWorkflowRunOutput } from "../runs/foreground/executor-outputs.js";
 import { isWorkflowDefinition, workflowDefinitionRequirementMessage } from "../runs/foreground/executor-child-helpers.js";
 import { getDurableBackend } from "../durable/factory.js";
+import { classifyReturnedRunStatus } from "./run-returned-status.js";
 import { createToolPrimitive, createCheckpointIdGenerator } from "../durable/tool-primitive.js";
 import { persistDurableCacheEntry } from "../durable/resume-catalog.js";
 import { createDurableStagePrimitive, createDurableTaskPrimitive, recordStageCheckpoint, createStageReplayKeyGenerator, recordCachedStageWithTracker } from "../durable/stage-primitive.js";
@@ -423,15 +424,16 @@ export async function run<
     assertWorkflowRunOutputs(def.name, result, def.outputs);
     assertWorkflowCreatedStage(runSnapshot);
     await durableBackend.flush?.();
-    const recorded = activeStore.recordRunEnd(runId, "completed", result);
-    appendRunEndWhenRecorded(opts.persistence, recorded, { runId, status: "completed", result, ts: Date.now() });
-    durableBackend.setWorkflowStatus(runId, "completed");
+    const returned = classifyReturnedRunStatus(result);
+    const recorded = activeStore.recordRunEnd(runId, returned.status, result, returned.error, returned.metadata);
+    appendRunEndWhenRecorded(opts.persistence, recorded, { runId, status: returned.status, result, ...(returned.error !== undefined ? { error: returned.error } : {}), ...(returned.metadata ?? {}), ts: Date.now() });
+    durableBackend.setWorkflowStatus(runId, returned.status, undefined, returned.metadata?.resumable);
     await durableBackend.flush?.();
     if (opts.persistence && durableBackend.persistent) {
       const cacheEntry = durableBackend.toCacheEntry(runId);
       if (cacheEntry) persistDurableCacheEntry(opts.persistence, cacheEntry);
     }
-    return reconcileTerminalRunResult(runId, runSnapshot, activeStore, { status: "completed", result }, opts.onRunEnd);
+    return reconcileTerminalRunResult(runId, runSnapshot, activeStore, { status: returned.status, result, error: returned.error }, opts.onRunEnd);
   } catch (err) {
     const selectedExit = findWorkflowExitSignal(err, exitScope) ?? findWorkflowExitSignal(ownController.signal.reason, exitScope);
     if (selectedExit !== undefined) return await finalizers.finalizeWorkflowExit(selectedExit);

package/dist/builtin/workflows/src/extension/config-file-loader.ts

@@ -68,7 +68,7 @@ function validateConfig(value: unknown): string | null {
       }
       for (const item of notifyOn) {
         if (!isWorkflowLifecycleNoticeKind(item)) {
-          return `"workflowNotifications.notifyOn" entries must be "completed", "failed", or "awaiting_input", got ${JSON.stringify(item)}`;
+          return `"workflowNotifications.notifyOn" entries must be "completed", "failed", "blocked", or "awaiting_input", got ${JSON.stringify(item)}`;
         }
       }
     }

package/dist/builtin/workflows/src/extension/config-loader.ts

@@ -182,7 +182,7 @@ export const WORKFLOW_CONFIG_DEFAULTS = {
   resumeInFlight: "ask" as const,
   workflowNotifications: {
     enabled: true,
-    notifyOn: ["completed", "failed", "awaiting_input"] as const,
+    notifyOn: ["completed", "failed", "blocked", "awaiting_input"] as const,
   },
 } as const;
 

package/dist/builtin/workflows/src/extension/lifecycle-notifications.ts

@@ -21,11 +21,12 @@ import { renderWorkflowNoticeCard, type WorkflowNoticeTone } from "../tui/workfl
 export const LIFECYCLE_NOTICE_CUSTOM_TYPE = "workflows:lifecycle-notice";
 export const LIFECYCLE_NOTICE_SNIPPET_LIMIT = 240;
 
-export type WorkflowLifecycleNoticeKind = "completed" | "failed" | "awaiting_input";
+export type WorkflowLifecycleNoticeKind = "completed" | "failed" | "blocked" | "awaiting_input";
 
 export const WORKFLOW_LIFECYCLE_NOTICE_KINDS = [
   "completed",
   "failed",
+  "blocked",
   "awaiting_input",
 ] as const satisfies readonly WorkflowLifecycleNoticeKind[];
 
@@ -95,8 +96,9 @@ export function seedWorkflowLifecycleNotificationState(
 ): void {
   for (const run of snapshot.runs) {
     if (!isTopLevelWorkflowRun(run)) continue;
-    if ((run.status === "completed" || run.status === "failed") && run.endedAt !== undefined) {
-      state.deliveredTerminalRuns.add(terminalRunKey(run.status, run.id));
+    const noticeKind = terminalNoticeKind(run);
+    if (noticeKind !== undefined && run.endedAt !== undefined) {
+      state.deliveredTerminalRuns.add(terminalRunKey(noticeKind, run.id));
     }
     if (run.pendingPrompt !== undefined) {
       state.deliveredInputPrompts.add(runAwaitingInputKey(run.id, run.pendingPrompt));
@@ -186,9 +188,10 @@ export function installWorkflowLifecycleNotifications(
 
   const emitTerminalNoticeOnce = (
     run: RunSnapshot,
-    kind: "completed" | "failed",
+    kind: "completed" | "failed" | "blocked",
   ): void => {
-    if (run.status !== kind || run.endedAt === undefined || !notifyOn.has(kind)) {
+    const noticeKind = terminalNoticeKind(run);
+    if (noticeKind !== kind || run.endedAt === undefined || !notifyOn.has(kind)) {
       return;
     }
 
@@ -231,6 +234,7 @@ export function installWorkflowLifecycleNotifications(
       if (!isTopLevelWorkflowRun(run)) continue;
       emitTerminalNoticeOnce(run, "completed");
       emitTerminalNoticeOnce(run, "failed");
+      emitTerminalNoticeOnce(run, "blocked");
 
       if (!notifyOn.has("awaiting_input")) continue;
       emitRunAwaitingInputNoticeOnce(run);
@@ -273,6 +277,10 @@ export function formatWorkflowLifecycleNoticeText(details: WorkflowLifecycleNoti
     const errorText = details.error ? `: ${details.error}` : "";
     return `✗ Workflow "${workflowName}" failed (run ${details.runId}${stageText})${errorText}. Inspect: /workflow status ${details.runId}`;
   }
+  if (details.kind === "blocked") {
+    const errorText = details.error ? `: ${details.error}` : "";
+    return `! Workflow "${workflowName}" ended blocked (run ${details.runId})${errorText}. Inspect: /workflow status ${details.runId}`;
+  }
   const prompt = details.promptMessage ? ` Prompt: ${details.promptMessage}` : "";
   if (details.scope === "run") {
     return `？ Workflow "${workflowName}" needs input (run ${details.runId}).${prompt} Respond: /workflow connect ${details.runId} to answer this run-level prompt.`;
@@ -286,18 +294,19 @@ export function formatWorkflowLifecycleNoticeText(details: WorkflowLifecycleNoti
 
 function makeTerminalNotice(
   run: RunSnapshot,
-  kind: "completed" | "failed",
+  kind: "completed" | "failed" | "blocked",
 ): WorkflowLifecycleNoticeDetails {
   const failedStage = run.failedStageId
     ? run.stages.find((stage) => stage.id === run.failedStageId)
     : undefined;
+  const error = run.error ?? (kind === "blocked" ? run.exitReason : undefined);
   return {
     kind,
     scope: "run",
     runId: run.id,
     workflowName: run.name,
     status: run.status,
-    ...(run.error ? { error: truncateSnippet(run.error) } : {}),
+    ...(error ? { error: truncateSnippet(error) } : {}),
     ...(run.failedStageId ? { failedStageId: run.failedStageId } : {}),
     ...(failedStage ? { stageId: failedStage.id, stageName: failedStage.name } : {}),
     ...(run.durationMs !== undefined ? { durationMs: run.durationMs } : {}),
@@ -320,7 +329,15 @@ function jsonString(value: string): string {
   return JSON.stringify(value);
 }
 
-function terminalRunKey(kind: "completed" | "failed", runId: string): string {
+function terminalNoticeKind(run: RunSnapshot): "completed" | "failed" | "blocked" | undefined {
+  if (run.status === "failed" || run.status === "blocked") return run.status;
+  if (run.status !== "completed") return undefined;
+  const returnedStatus = run.result?.["status"];
+  if (returnedStatus === "failed" || returnedStatus === "blocked") return returnedStatus;
+  return "completed";
+}
+
+function terminalRunKey(kind: "completed" | "failed" | "blocked", runId: string): string {
   return `${kind}:${runId}`;
 }
 
@@ -361,19 +378,27 @@ function renderLifecycleNoticeCard(
   details: WorkflowLifecycleNoticeDetails,
   opts: { width: number; theme?: GraphTheme; fallbackText: string },
 ): string[] {
-  const tone: WorkflowNoticeTone = details.kind === "failed" ? "error" : details.kind === "awaiting_input" ? "warning" : "success";
+  const tone: WorkflowNoticeTone = details.kind === "failed"
+    ? "error"
+    : details.kind === "awaiting_input" || details.kind === "blocked"
+      ? "warning"
+      : "success";
   const title = details.kind === "failed"
     ? "WORKFLOW FAILED"
     : details.kind === "awaiting_input"
       ? "WORKFLOW INPUT"
-      : "WORKFLOW COMPLETE";
-  const glyph = details.kind === "failed" ? "✗" : details.kind === "awaiting_input" ? "？" : "✓";
+      : details.kind === "blocked"
+        ? "WORKFLOW BLOCKED"
+        : "WORKFLOW COMPLETE";
+  const glyph = details.kind === "failed" ? "✗" : details.kind === "awaiting_input" ? "？" : details.kind === "blocked" ? "!" : "✓";
   const stage = details.stageName ?? details.failedStageId ?? details.stageId;
   const headline = details.kind === "failed"
     ? `Workflow "${details.workflowName}" failed`
     : details.kind === "awaiting_input"
       ? `Workflow "${details.workflowName}" needs input`
-      : `Workflow "${details.workflowName}" completed`;
+      : details.kind === "blocked"
+        ? `Workflow "${details.workflowName}" ended blocked`
+        : `Workflow "${details.workflowName}" completed`;
   return renderWorkflowNoticeCard({
     title,
     glyph,

package/dist/builtin/workflows/src/shared/persistence-restore-helpers.ts

@@ -335,7 +335,7 @@ export function restoreTerminalRuns(entries: readonly SessionEntry[], store: Sto
     const exitReason = end["exitReason"];
     const resumable = end["resumable"];
     const restoredAuthorExit = isWorkflowExitTerminalStatus(status) &&
-      (exited === true || status !== "completed" || typeof exitReason === "string" || resumable === false);
+      (exited === true || typeof exitReason === "string");
     if (status === "completed" && !restoredAuthorExit && stages.some((stage) => stage.status !== "completed")) continue;
     store.recordRunStart({
       id: runId,

package/dist/builtin/workflows/src/shared/store-public-types.ts

@@ -83,8 +83,9 @@ export interface Store {
    * Records the end of a run.
    * Returns `true` if state changed, `false` if the run was not found or
    * already in a terminal state (completed | failed | killed | skipped | cancelled | blocked).
-   * `result` is applied for intentional success/exit statuses (completed | skipped | cancelled | blocked).
-   * `error` is only applied for status "failed" | "killed".
+   * `result` is applied for intentional success/exit statuses (completed | skipped | cancelled | blocked)
+   * and for workflows that intentionally return a failed status with structured outputs.
+   * `error` is applied for status "failed" | "killed" | "blocked".
    */
   recordRunEnd(
     runId: string,

package/dist/builtin/workflows/src/shared/store-run-methods.ts

@@ -82,11 +82,12 @@ export function createRunStoreMethods(context: StoreContext): RunStoreMethods {
         run.pausedAt = undefined;
       }
       run.durationMs = elapsedRunMs(run, run.endedAt);
+      if (result !== undefined && shouldStoreRunResult(status)) run.result = result;
       const wasBlocked = run.blockedAt !== undefined || run.failureDisposition === "active_blocked";
       delete run.blockedAt;
       if (status === "completed" || status === "skipped" || status === "cancelled" || status === "blocked") {
-        if (result !== undefined) run.result = result;
         clearRunFailureMetadata(run);
+        if (status === "blocked" && error !== undefined) run.error = error;
         if (metadata !== undefined) applyRunEndMetadata(run, metadata);
       } else {
         if (wasBlocked && error === undefined) delete run.error;
@@ -227,3 +228,7 @@ export function createRunStoreMethods(context: StoreContext): RunStoreMethods {
     },
   };
 }
+
+function shouldStoreRunResult(status: RunStatus): boolean {
+  return status === "completed" || status === "skipped" || status === "cancelled" || status === "blocked" || status === "failed";
+}

package/dist/core/http-dispatcher.d.ts

@@ -1,6 +1,4 @@
 export declare const DEFAULT_HTTP_IDLE_TIMEOUT_MS = 600000;
-/** Connect-phase timeout so a black-holed/firewalled host fails fast instead of hanging until the OS TCP timeout. */
-export declare const HTTP_CONNECT_TIMEOUT_MS = 10000;
 export declare const HTTP_IDLE_TIMEOUT_CHOICES: readonly [{
     readonly label: "30 sec";
     readonly timeoutMs: 30000;
@@ -28,7 +26,9 @@ export declare function formatHttpIdleTimeoutMs(timeoutMs: number): string;
  * Keep HTTP/2 disabled for now because some Node/undici combinations have
  * produced stream-reset crashes, and use a configurable idle timeout so stale
  * connections are eventually reclaimed while long-running requests remain
- * supported.
+ * supported. Do not install a fixed connect-phase timeout here: under Pier and
+ * other policy/proxy layers, CONNECT establishment can be slower than normal
+ * internet egress and should surface through the provider/agent retry path.
  */
 export declare function configureHttpDispatcher(timeoutMs?: number): void;
 //# sourceMappingURL=http-dispatcher.d.ts.map