@bastani/atomic 0.8.21-0 → 0.8.22-0

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

@@ -1,6 +1,6 @@
 {
   "name": "@bastani/web-access",
-  "version": "0.8.21-0",
+  "version": "0.8.22-0",
   "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

@@ -6,24 +6,90 @@ The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.1.0/),
 
 ## [Unreleased]
 
+## [0.8.22-0] - 2026-06-01
+
+### Breaking Changes
+
+- Migrated workflow input/output declarations to a TypeBox-native schema system, replacing the legacy `{ type, required, default, choices }` descriptor. Authors now declare inputs/outputs with TypeBox schemas (`.input("prompt", Type.String({ description }))`, `.input("count", Type.Number({ default: 2 }))`, `.input("flavor", Type.Union([Type.Literal("a"), Type.Literal("b")]))`, `.output("packet", Type.Object({ topic: Type.String(), score: Type.Number() }))`). `ctx.inputs`, the `.run()` return, and `ctx.workflow(child).outputs` are precisely typed via `Static<>`, and the runtime validates inputs and outputs with TypeBox `Value`. `Type` is re-exported from `@bastani/workflows` (with types `Static` and `TSchema`) so workflow authors single-import. An optional field is declared with `Type.Optional(...)`; a `default` keeps the key required at the type level. This is a clean break with no legacy descriptor support.
+- Removed the unshipped `.import(...)` workflow builder API and string-alias child workflow calls. Workflows now compose children by importing compiled workflow definitions with TypeScript imports and passing those definitions directly to `ctx.workflow(workflowDefinition, options)`.
+- Removed parent-side child output selection/renaming from `ctx.workflow(...)`. Parent workflows now receive the child's declared `.output(...)` contract as `child.outputs`.
+- Made workflow outputs fully explicit so a workflow's input/output contract is the stable, self-documenting surface other workflows compose against. A workflow now exposes exactly the outputs it declares with `.output(...)`, and a `.run()` return that contains a key the workflow did not declare fails the run with `atomic-workflows: ... returned undeclared output "<key>"`. The implicit string `result` output and the `child.rawOutput` escape hatch were removed: declare `.output("result", schema)` and return `{ result }` if you want a `result` output, and declare every other field a parent should read.
+- Changed default transcript inspection from inlining about 50 recent entries to a reference-first 5-entry preview with `sessionFile`/`transcriptPath`; pass explicit `tail` or `limit` to override the preview size.
+- Removed the obsolete `.humanInTheLoop(...)` workflow builder API and its declaration-time interaction metadata; workflows should use runtime `ctx.ui.*` calls directly when they need human input.
+
+### Added
+
+- Added TypeBox as the workflow schema engine and re-exported `Type` (plus the `Static` and `TSchema` types) from the `@bastani/workflows` public entry point so authors single-import the authoring surface.
+- Added TypeScript module-style workflow composition: parent workflows can pass compiled child workflow definitions directly to `ctx.workflow(compiledWorkflow, options)`. The bundled `deep-research-codebase`, `goal`, `ralph`, and `open-claude-design` workflows are reusable from `@bastani/workflows/builtin` or individual builtin module paths.
+- Added first-class workflow composition: child workflows can declare `.output()` contracts, and `ctx.workflow()` runs compiled child workflow definitions as nested runs with input validation, schema-validated declared outputs, and the nested workflow's stages flattened inline into the parent graph ([#1071](https://github.com/bastani-inc/atomic/issues/1071)).
+
+### Removed
+
+- Removed the `zod` dependency from `@bastani/workflows`; JSON-serializable validation and input/output validation now run on TypeBox `Value`.
+
+### Changed
+
+- Raised the bundled `deep-research-codebase` default `max_concurrency` from 4 to 100 so partitioned research stages fan out in parallel by default; lower it via the `max_concurrency` input on rate-limited or cost-sensitive setups (with `max_partitions` up to 100 and one locator/pattern-finder/analyzer/online-researcher specialist per partition, a single run can otherwise spawn many concurrent stage sessions).
+- Declared explicit output contracts for the bundled `deep-research-codebase`, `goal`, `ralph`, and `open-claude-design` workflows so parent workflows receive schema-validated child outputs.
+- Renamed workflow runtime error prefixes from `pi-workflows:` to `atomic-workflows:`.
+- Render workflow stage notices, lifecycle notices, and HiL answer notices as compact emoji-free TUI cards instead of plain wrapped rows, matching the debugger warning card treatment.
+- Flatten imported workflows in the expanded run graph so a `ctx.workflow()` import reads as a single flat layout: the child run's stages stand in for the import instead of rendering an extra boundary "information" node above them. The boundary's incoming dependencies become the child roots' dependencies and downstream stages rewire to the child's terminal stages, so a depth-1 composition of single-stage leaves now shows one node per import instead of two. A boundary whose child run produced no stages of its own is still kept as a single node so the import stays visible. Affects the graph overlay, session list, `status` stage counts/details, and attach resolution alike.
+- Updated the registered `workflow` tool description so agent-facing metadata covers named and direct runs, discovery, inspection, prompt answering/steering, run control, and reload ([#1151](https://github.com/bastani-inc/atomic/issues/1151)).
+
+### Fixed
+
+- Fixed the programmatic named-workflow runner (`runWorkflow`) to resolve declared input defaults before validating, so an input declared with both `required: true` and a `default` is no longer rejected as missing when omitted, matching the resolve-then-validate order used by the other dispatch paths.
+- Fixed slow workflow discovery outside the compiled binary by always resolving the `@bastani/workflows` SDK (and its builtin submodules) to in-memory virtual modules in the jiti loader instead of aliasing to the on-disk package. The alias path re-evaluated the entire SDK module graph once per workflow file (the loader keeps `moduleCache: false` so edits stay observable), so projects with many workflow files saw multi-second discovery and timing-sensitive workflow tests flaked at the default test timeout; workflow files are still evaluated fresh from disk, so `/workflow reload` keeps observing edits.
+- Fixed headless workflow shutdown so completed stage chat handles are disposed on CLI quit, allowing successful non-interactive `/workflow` runs to return to the shell prompt ([#1167](https://github.com/bastani-inc/atomic/issues/1167)).
+- Fixed bundled workflow modules so they import the leaf `defineWorkflow` authoring API instead of the public package entrypoint, avoiding an ESM initialization cycle during builtin workflow discovery.
+- Fixed workflow HIL prompts so awaiting-input and answered-prompt notices no longer wake the main chat agent, answer notices stay out of LLM context and explicitly forbid auto-answering later HIL prompts, and stale Enter input from graph/connect focus transitions is quarantined when prompts appear after the attach pane is already open, covering input, confirm, select, and editor prompts ([#1163](https://github.com/bastani-inc/atomic/issues/1163)).
+- Fixed workflow run widgets so store changes and elapsed-time ticks repaint through a long-lived reactive widget instead of waiting for unrelated chat input or remounting the widget ([#1150](https://github.com/bastani-inc/atomic/issues/1150)).
+- Fixed `/workflow reload` and `workflow({ action: "reload" })` so newly added package-manifest workflow entries are rediscovered in-process without requiring top-level `/reload` or restart ([#1155](https://github.com/bastani-inc/atomic/issues/1155)).
+- Fixed completed child workflow boundary stages rendering as empty graph nodes by showing the child workflow name, child run id prefix, and output count in the node card.
+- Fixed the background workflow widget (`BACKGROUND` panel) listing nested `ctx.workflow()` child runs as separate top-level entries, so a two-level composition no longer shows the root, parent, and child runs as three separate items with an inflated `N runs` count. The widget now applies the same top-level visibility rule (`run.parentRunId === undefined`) already used by `statusRuns`, the `status` action, and the `/workflow connect` session picker. The "needs attention" (awaiting HiL input) badge still fires for a hidden nested descendant by attributing its awaiting state to the visible top-level ancestor via `rootRunId`, so a HiL prompt waiting inside an imported child workflow stays discoverable.
+- Fixed the bundled deep-research workflow line-count heuristic hanging Windows CI by replacing its POSIX `wc` subprocess with portable in-process line counting.
+- Kept workflow stage model metadata as the raw model id while surfacing Codex fast mode as a separate visible `fast` marker on workflow node cards, including stages that use the default Atomic SDK adapter settings manager.
+- Show workflow Codex fast-mode metadata on running workflow nodes as soon as the stage session starts, including explicit-model stages, catalog-resolved bare model aliases, and custom resource-loader stages whose settings manager is created by the Atomic SDK.
+- Kept workflow Codex fast-mode markers synchronized across fallback attempts: running nodes now update when fallback switches to a fast-eligible model, completed nodes clear stale `fast` markers when fallback finishes on a non-eligible model, and prompt-adapter stages no longer create SDK sessions only to compute fast metadata.
+- Made successful headless `/workflow` informational commands emit displayable command output instead of dropping success messages through the no-op non-interactive UI reporter, made degraded terminal-detail fallback rendering avoid fabricating successful status for unknown run states, and preserved non-interactive execution policy when resuming failed workflow runs ([#1123](https://github.com/bastani-inc/atomic/issues/1123)).
+- Suppressed workflow lifecycle steer notices while awaited non-interactive workflow dispatch is already waiting for terminal completion, and made workflow chat-surface custom-message content printable by default for headless list/status/detail/dispatch/killed output ([#1123](https://github.com/bastani-inc/atomic/issues/1123)).
+- Re-enabled deterministic non-interactive workflow execution: headless sessions keep the `workflow` tool, `/workflow <name> key=value` skips interactive pickers, named workflow dispatch waits for the terminal run snapshot, declared human-in-the-loop workflows are rejected, top-level `ctx.ui.*` is unavailable, and non-interactive stage sessions exclude `ask_user_question` without binding broker-backed extension UI ([#1123](https://github.com/bastani-inc/atomic/issues/1123)).
+- Added a concise display-only `User responded with: ...` workflow HiL answer notice so the main chat transcript records the answer for user-visible audit without triggering an autonomous model turn or adding the notice to LLM context ([#1137](https://github.com/bastani-inc/atomic/issues/1137)).
+- Suppressed workflow HiL answer interrupt notices for prompts answered by the main-chat `workflow` tool, so tool-driven answers do not dismiss the current main-chat turn with workflow-chat wording ([#1137](https://github.com/bastani-inc/atomic/issues/1137)).
+- Hardened workflow awaiting-input lifecycle notices and readiness-gate answers: awaiting-input lifecycle states now emit main-chat steer notices for dedupe-aware model visibility while preserving the existing `/workflow connect` and `workflow send` response hints, readiness gates use unique prompt ids per gate instance, and duplicate/raced `workflow send` answers for a just-resolved brokered input request now report the request as already answered instead of falling through to `No matching pending prompt` ([#1137](https://github.com/bastani-inc/atomic/issues/1137)).
+- Prevented Enter used for `/workflow connect`, graph attach, direct stage attach, unrelated graph-node inspection, or transition key-repeat from immediately submitting workflow HIL prompts; prompt submission now requires the user to be past the attach transition and press Enter while the specific prompted node is attached, leaving sibling HIL prompts pending ([#1148](https://github.com/bastani-inc/atomic/issues/1148)).
+- Fixed completed human-in-the-loop prompt archives so reattached read-only prompt nodes keep their prompt card scrollable with keyboard and mouse-wheel input ([#1140](https://github.com/bastani-inc/atomic/issues/1140)).
+- Kept the workflow graph overlay navigable while a stage-local human-input prompt is awaiting a response: graph navigation, the stage switcher, Ctrl+D detach, and mouse-wheel scrolling now stay owned by graph mode until the user explicitly attaches to the prompted stage ([#1141](https://github.com/bastani-inc/atomic/issues/1141)).
+- Preserved literal slash text entry in legacy run-level workflow prompt cards, so paths and URLs such as `/tmp/file` remain typable while the prompt owns input ([#1141](https://github.com/bastani-inc/atomic/issues/1141)).
+- Cleared brokered stage HIL UI during terminal stage cleanup so a completed workflow does not leave an active `ask_user_question`/custom prompt card visible after detach ([#1141](https://github.com/bastani-inc/atomic/issues/1141)).
+- Fixed workflow HIL select/editor prompts so navigation, scroll input, and ambiguous Escape-prefix input cannot advance prompts before an explicit configured confirm/Enter submit or Ctrl+C skip. Graph-mode prompt cards now honor custom select keybindings, and Enter submitted during connect/attach/detach transitions is time-boxed so key repeat cannot attach or answer the next prompt accidentally ([#1148](https://github.com/bastani-inc/atomic/issues/1148)).
+- Preserved synthetic workflow HiL prompt nodes as `awaiting_input` when an unrelated main-chat `ask_user_question` result arrives, so choosing to leave a workflow prompt open no longer makes the orchestrator node look like it resumed while the prompt is still unanswered ([#1137](https://github.com/bastani-inc/atomic/issues/1137)).
+- Fixed headless `/workflow <name>` print-mode automation so successful non-interactive runs wait for terminal snapshots and emit printable run-detail summaries, while terminal failures surface as command-visible errors instead of silent zero-output successes ([#1156](https://github.com/bastani-inc/atomic/issues/1156)).
+
+## [0.8.21] - 2026-05-30
+
+### Changed
+
+- Promoted the 0.8.21 prerelease changes to a stable release.
+
 ## [0.8.21-0] - 2026-05-30
 
 ### Added
 
 - Enabled the `/workflow send` action (with `delivery: "answer"`) to answer a stage's brokered structured prompts — an in-stage `ask_user_question` call or the deterministic readiness gate — without the interactive TUI/graph overlay. These prompts resolve through the `StageUiBroker` (`ctx.ui.custom`) rather than the simple `store.pendingPrompt` model, so `send` previously no-op'd ("No pending prompt to answer") and they could only be answered by attaching the graph viewer. The stage UI broker now carries an optional headless-answer adapter (`provideStagePrompt` / `peekStagePrompt` / `answerStagePrompt` / `clearStagePrompt`), keyed by `(runId, stageId)`, that maps a simple answer — a free-text reply, an option label, a 1-based option index, comma-separated labels for multi-select, or a pre-built `response` — into the `QuestionnaireResult` the tool expects. The executor's existing `ask_user_question` watcher captures the tool-call `args` to build the adapter, and the readiness gate registers one statically; both surface a serializable `inputRequest` descriptor on the stage snapshot (shown in `/workflow stage` / `stages` output) so an orchestrating agent can see the questions/options and answer them programmatically.
-- Added a deterministic readiness gate after `ask_user_question` tool calls inside workflow stages. When a stage's model turn issues an `ask_user_question` call, the workflow re-uses the structured `ask_user_question` UI — rendered inline in the attached stage chat via the stage UI broker — to ask "Are you ready to move on to the next stage?" (Yes/No) before completing/advancing the stage. Answering **No** (or "Chat about this"/cancel) steers the stage with a continuation message so the conversation keeps going, then re-shows the gate after the next turn, repeating until **Yes**; **Yes** resumes normal sequential and dependent-parallel progression. Detection is independent of how the underlying UI is implemented, and the executor exposes a `confirmStageReadiness` seam so the gate is fully testable. `@bastani/atomic` now also exports `createAskUserQuestionToolDefinition` so first-party extensions can invoke the structured prompt deterministically ([#1099](https://github.com/flora131/atomic/issues/1099)).
+- Added a deterministic readiness gate after `ask_user_question` tool calls inside workflow stages. When a stage's model turn issues an `ask_user_question` call, the workflow re-uses the structured `ask_user_question` UI — rendered inline in the attached stage chat via the stage UI broker — to ask "Are you ready to move on to the next stage?" (Yes/No) before completing/advancing the stage. Answering **No** (or "Chat about this"/cancel) steers the stage with a continuation message so the conversation keeps going, then re-shows the gate after the next turn, repeating until **Yes**; **Yes** resumes normal sequential and dependent-parallel progression. Detection is independent of how the underlying UI is implemented, and the executor exposes a `confirmStageReadiness` seam so the gate is fully testable. `@bastani/atomic` now also exports `createAskUserQuestionToolDefinition` so first-party extensions can invoke the structured prompt deterministically ([#1099](https://github.com/bastani-inc/atomic/issues/1099)).
 
 ### Fixed
 
 - Fixed background workflow HIL prompt nodes becoming input-dead when they appeared while the orchestrator graph overlay was already open: the visible overlay now reclaims focus on awaiting-input store updates, and the graph auto-focuses newly awaiting prompt nodes so Enter attaches directly to the HIL response UI.
 - Fixed the `/workflow status <id>` run-detail card still ticking (and flickering the whole screen) even after the chat-surface clock was frozen: a running stage's in-flight tool-activity label (e.g. `bash · 6s`) was computed inside `stageActivityString()` from a fresh `Date.now()`, bypassing the capture-once `now` that `renderRunDetail` threads everywhere else. While a background workflow ran, the below-editor companion widget drove a full re-render roughly once per second (and on every store mutation); each re-render advanced that label, and once the detail card had scrolled above the viewport fold pi-tui took the full-screen-clear path (CSI 2J + CSI H + CSI 3J), read as whole-page + chat-box flicker. `stageActivityString()` now uses the threaded `now`, so scrollback run-detail cards are byte-stable across re-renders (the companion widget still owns the live, ticking view).
 - Fixed the `/workflow status` and `/workflow status <id>` chat-surface cards triggering a full-screen redraw (clear + scrollback wipe) every render tick once they scroll above the viewport fold. Their custom-message renderer's `render()` lambda re-runs on every TUI frame (pi-tui fans out to every child each `doRender`, and the live workflow/subagent widgets request a frame ~12×/sec while runs are active), and it called `renderStatusList` / `renderRunDetail` without a captured clock, so they fell through to `Date.now()` each frame and ticked the `elapsed` / `running` labels — and pi-tui full-clears the screen (CSI 2J + CSI H + CSI 3J) whenever a changed line sits above the fold, read as whole-screen flicker on terminals without synchronized output (e.g. mosh). The renderer now captures wall-clock once when the chat entry's component is created and threads it through to the status/detail renderers, so these point-in-time scrollback snapshots stay byte-stable across re-renders (the live widget still owns live state). Mirrors the earlier tool-result status-card fix.
-- Fixed answering a stage's brokered structured prompt (an in-stage `ask_user_question` or the readiness gate) via `/workflow send` replying **"User declined to answer questions"** and leaving the stage stuck `running` with no pending input — even though a real answer was sent. A structured `response` was forwarded verbatim to the `ask_user_question` tool: a JSON-encoded string was captured as raw free text, and — most damagingly — any object with an `answers` array (e.g. the orchestrator-friendly `{ answers: [{ question, answer }] }`) was treated as a pre-built `QuestionnaireResult` despite lacking the numeric `questionIndex` the tool envelope requires, so it matched no question segment and declined. `coerceStageInputAnswer` now forwards `raw` ONLY for a fully-formed result (every answer carries a numeric `questionIndex`); every other shape — JSON strings, `answers[]`/`questions[]` entries, flat `{ answer | label | selected }`, string arrays — is normalized to a label/index/multi-select value that the stage prompt adapter resolves against the question's options (assigning the correct `questionIndex`). The readiness decision (`readinessResultMeansAdvance`) matches the advance option case/whitespace-insensitively and via `selected[]`. A new regression test drives the real `StageUiBroker` + `ask_user_question` tool resolution path (the seam the earlier unit tests missed) and reproduces the decline for the `answers[]`-without-`questionIndex` payload. The "keep exploring"/stay path is unchanged ([#1099](https://github.com/flora131/atomic/issues/1099)).
-- Hardened the `/workflow` tool-result renderer against missing or unshaped payloads. `renderResult` is passed `result.details`, which can be absent during streaming/partial renders or on error paths, and previously dereferenced a missing `action` and crashed the TUI render loop. It now degrades gracefully (renders nothing for partials, a generic notice otherwise) ([#1099](https://github.com/flora131/atomic/issues/1099)).
-- Made `terminate: true` tool results deterministic when deriving a stage/task's result text: when an agent turn ends on a terminating tool, the stage runner now returns that tool result's output instead of any prose the model emitted before the tool call. This fixes the `goal` and `ralph` review gates, whose terminating `review_decision` structured-output tool emits clean JSON — previously the preceding assistant narration was captured instead, so the strict `JSON.parse` failed and a valid verdict was misclassified as a reviewer failure ("returned invalid structured JSON"), blocking quorum and forcing extra turns ([#1099](https://github.com/flora131/atomic/issues/1099)).
-- Tightened terminating-tool result detection when deriving a stage/task's result text. The stage runner now treats a trailing tool result as the deterministic turn output ONLY when that tool call actually returned `terminate: true` — tracked at runtime from the session's `tool_execution_end` events, since the tool-result message itself does not carry the flag — instead of whenever the last conversational message merely happened to be a tool result. A turn that ends on a non-terminating tool result (e.g. aborted or interrupted right after a tool call) now correctly falls back to the last assistant message, and that fallback no longer surfaces non-terminating tool output. Verified end-to-end against real `terminate: true`, `terminate: false`, and no-tool stages ([#1099](https://github.com/flora131/atomic/issues/1099)).
-- Fixed a mid-turn `ask_user_question` (or readiness gate) rendering but being input-dead in the attached stage chat. After a readiness-gate "stay" returns to the composer and the user submits another message, the agent's next question mounts while host keyboard focus has drifted off the overlay; the focus re-assertion was suppressed during streaming (a blunt guard added to avoid a prior continuation stall), so arrows/Enter were ignored and nothing could be answered or sent to the model. The overlay's `requestFocus` is now idempotent — it grabs focus only when the overlay does not already own it (`isFocused()`), so the redundant `focus()` that stalled the stream never runs. The stage chat now requests focus whenever a custom UI is shown (including mid-turn) and the focus-hold timer keeps a mounted question focused, trusting `requestFocus` to no-op when unnecessary ([#1120](https://github.com/flora131/atomic/issues/1120)).
-- Detaching or closing the attached stage chat (Ctrl+D / Ctrl+C) while a stage has a pending human-input request — an agent `ask_user_question` call or the readiness gate — no longer cancels that request. Previously, tearing down the chat view (or its stage UI broker host) rejected the pending prompt, so the stage silently dropped out of `awaiting_input` and the question disappeared. Detaching, closing, and disposing now only stop *displaying* the request: it stays pending (the stage remains `awaiting_input`) and is re-displayed when a host re-attaches. A pending request is settled only by the user answering (broker resolve) or the run aborting via its `AbortSignal` — the single chokepoints for ending a human-input request ([#1099](https://github.com/flora131/atomic/issues/1099)).
+- Fixed answering a stage's brokered structured prompt (an in-stage `ask_user_question` or the readiness gate) via `/workflow send` replying **"User declined to answer questions"** and leaving the stage stuck `running` with no pending input — even though a real answer was sent. A structured `response` was forwarded verbatim to the `ask_user_question` tool: a JSON-encoded string was captured as raw free text, and — most damagingly — any object with an `answers` array (e.g. the orchestrator-friendly `{ answers: [{ question, answer }] }`) was treated as a pre-built `QuestionnaireResult` despite lacking the numeric `questionIndex` the tool envelope requires, so it matched no question segment and declined. `coerceStageInputAnswer` now forwards `raw` ONLY for a fully-formed result (every answer carries a numeric `questionIndex`); every other shape — JSON strings, `answers[]`/`questions[]` entries, flat `{ answer | label | selected }`, string arrays — is normalized to a label/index/multi-select value that the stage prompt adapter resolves against the question's options (assigning the correct `questionIndex`). The readiness decision (`readinessResultMeansAdvance`) matches the advance option case/whitespace-insensitively and via `selected[]`. A new regression test drives the real `StageUiBroker` + `ask_user_question` tool resolution path (the seam the earlier unit tests missed) and reproduces the decline for the `answers[]`-without-`questionIndex` payload. The "keep exploring"/stay path is unchanged ([#1099](https://github.com/bastani-inc/atomic/issues/1099)).
+- Hardened the `/workflow` tool-result renderer against missing or unshaped payloads. `renderResult` is passed `result.details`, which can be absent during streaming/partial renders or on error paths, and previously dereferenced a missing `action` and crashed the TUI render loop. It now degrades gracefully (renders nothing for partials, a generic notice otherwise) ([#1099](https://github.com/bastani-inc/atomic/issues/1099)).
+- Made `terminate: true` tool results deterministic when deriving a stage/task's result text: when an agent turn ends on a terminating tool, the stage runner now returns that tool result's output instead of any prose the model emitted before the tool call. This fixes the `goal` and `ralph` review gates, whose terminating `review_decision` structured-output tool emits clean JSON — previously the preceding assistant narration was captured instead, so the strict `JSON.parse` failed and a valid verdict was misclassified as a reviewer failure ("returned invalid structured JSON"), blocking quorum and forcing extra turns ([#1099](https://github.com/bastani-inc/atomic/issues/1099)).
+- Tightened terminating-tool result detection when deriving a stage/task's result text. The stage runner now treats a trailing tool result as the deterministic turn output ONLY when that tool call actually returned `terminate: true` — tracked at runtime from the session's `tool_execution_end` events, since the tool-result message itself does not carry the flag — instead of whenever the last conversational message merely happened to be a tool result. A turn that ends on a non-terminating tool result (e.g. aborted or interrupted right after a tool call) now correctly falls back to the last assistant message, and that fallback no longer surfaces non-terminating tool output. Verified end-to-end against real `terminate: true`, `terminate: false`, and no-tool stages ([#1099](https://github.com/bastani-inc/atomic/issues/1099)).
+- Fixed a mid-turn `ask_user_question` (or readiness gate) rendering but being input-dead in the attached stage chat. After a readiness-gate "stay" returns to the composer and the user submits another message, the agent's next question mounts while host keyboard focus has drifted off the overlay; the focus re-assertion was suppressed during streaming (a blunt guard added to avoid a prior continuation stall), so arrows/Enter were ignored and nothing could be answered or sent to the model. The overlay's `requestFocus` is now idempotent — it grabs focus only when the overlay does not already own it (`isFocused()`), so the redundant `focus()` that stalled the stream never runs. The stage chat now requests focus whenever a custom UI is shown (including mid-turn) and the focus-hold timer keeps a mounted question focused, trusting `requestFocus` to no-op when unnecessary ([#1120](https://github.com/bastani-inc/atomic/issues/1120)).
+- Detaching or closing the attached stage chat (Ctrl+D / Ctrl+C) while a stage has a pending human-input request — an agent `ask_user_question` call or the readiness gate — no longer cancels that request. Previously, tearing down the chat view (or its stage UI broker host) rejected the pending prompt, so the stage silently dropped out of `awaiting_input` and the question disappeared. Detaching, closing, and disposing now only stop *displaying* the request: it stays pending (the stage remains `awaiting_input`) and is re-displayed when a host re-attaches. A pending request is settled only by the user answering (broker resolve) or the run aborting via its `AbortSignal` — the single chokepoints for ending a human-input request ([#1099](https://github.com/bastani-inc/atomic/issues/1099)).
 
 ## [0.8.20] - 2026-05-29
 
@@ -35,20 +101,20 @@ The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.1.0/),
 
 ### Added
 
-- Added main-chat lifecycle steer notices for workflow completion, failure, and awaiting-input pauses with global notification config controls ([#1085](https://github.com/flora131/atomic/issues/1085)).
+- Added main-chat lifecycle steer notices for workflow completion, failure, and awaiting-input pauses with global notification config controls ([#1085](https://github.com/bastani-inc/atomic/issues/1085)).
 
 ### Fixed
 
-- Fixed the background-workflow companion counter widget flickering every second while a run is active by updating a single long-lived above-editor widget in place instead of disposing and re-mounting a fresh widget factory on each elapsed-clock tick ([#1109](https://github.com/flora131/atomic/issues/1109)).
-- Fixed workflow lifecycle completion/failure/awaiting-input notices crashing the TUI on narrow or freshly-resized terminals: the notice component now wraps to the render width (hard-breaking long run ids) instead of emitting a single fixed line, which pi-tui rejects with a hard "Rendered line exceeds terminal width" throw ([#1109](https://github.com/flora131/atomic/issues/1109)).
-- Fixed the workflow companion counter widget triggering a full-screen redraw (clear + scrollback wipe) on every elapsed-clock tick after the terminal is resized while a run is active. The widget now mounts `belowEditor` instead of `aboveEditor`, keeping its live clock line within the bottom viewport; pi-tui full-clears the screen whenever a changed line sits above the viewport fold, and an above-editor widget was pushed above the fold once the editor/status region grew tall ([#1109](https://github.com/flora131/atomic/issues/1109)).
-- Disabled workflows in non-interactive (`-p` / `--print` / `--mode json`) sessions, which bind a no-op UI surface and cannot drive workflow prompts, pickers, or the graph overlay: the `workflow` tool is removed from the model's active tool set at session start, and the `/workflow` command (reachable via `atomic -p "/workflow …"`) is refused — preventing headless runs from stalling on work that can never complete ([#1096](https://github.com/flora131/atomic/issues/1096)).
-- Escaped workflow lifecycle notice text and structured response hints, isolated lifecycle send failures from store subscribers, and rejected empty lifecycle notification event lists ([#1085](https://github.com/flora131/atomic/issues/1085)).
-- Fixed stage awaiting-input lifecycle notice dedupe so promptless pauses after resolved prompts are not suppressed by historical prompt metadata ([#1085](https://github.com/flora131/atomic/issues/1085)).
-- Reset workflow lifecycle-notification dedupe state at chat session boundaries so reused workflow run IDs in later sessions still emit completion/failure/input notices ([#1085](https://github.com/flora131/atomic/issues/1085)).
-- Warn before starting or resuming another session when workflows are still in flight, allowing users to cancel before those runs are killed and current-session workflow history is cleared ([#1082](https://github.com/flora131/atomic/issues/1082)).
+- Fixed the background-workflow companion counter widget flickering every second while a run is active by updating a single long-lived above-editor widget in place instead of disposing and re-mounting a fresh widget factory on each elapsed-clock tick ([#1109](https://github.com/bastani-inc/atomic/issues/1109)).
+- Fixed workflow lifecycle completion/failure/awaiting-input notices crashing the TUI on narrow or freshly-resized terminals: the notice component now wraps to the render width (hard-breaking long run ids) instead of emitting a single fixed line, which pi-tui rejects with a hard "Rendered line exceeds terminal width" throw ([#1109](https://github.com/bastani-inc/atomic/issues/1109)).
+- Fixed the workflow companion counter widget triggering a full-screen redraw (clear + scrollback wipe) on every elapsed-clock tick after the terminal is resized while a run is active. The widget now mounts `belowEditor` instead of `aboveEditor`, keeping its live clock line within the bottom viewport; pi-tui full-clears the screen whenever a changed line sits above the viewport fold, and an above-editor widget was pushed above the fold once the editor/status region grew tall ([#1109](https://github.com/bastani-inc/atomic/issues/1109)).
+- Disabled workflows in non-interactive (`-p` / `--print` / `--mode json`) sessions, which bind a no-op UI surface and cannot drive workflow prompts, pickers, or the graph overlay: the `workflow` tool is removed from the model's active tool set at session start, and the `/workflow` command (reachable via `atomic -p "/workflow …"`) is refused — preventing headless runs from stalling on work that can never complete ([#1096](https://github.com/bastani-inc/atomic/issues/1096)).
+- Escaped workflow lifecycle notice text and structured response hints, isolated lifecycle send failures from store subscribers, and rejected empty lifecycle notification event lists ([#1085](https://github.com/bastani-inc/atomic/issues/1085)).
+- Fixed stage awaiting-input lifecycle notice dedupe so promptless pauses after resolved prompts are not suppressed by historical prompt metadata ([#1085](https://github.com/bastani-inc/atomic/issues/1085)).
+- Reset workflow lifecycle-notification dedupe state at chat session boundaries so reused workflow run IDs in later sessions still emit completion/failure/input notices ([#1085](https://github.com/bastani-inc/atomic/issues/1085)).
+- Warn before starting or resuming another session when workflows are still in flight, allowing users to cancel before those runs are killed and current-session workflow history is cleared ([#1082](https://github.com/bastani-inc/atomic/issues/1082)).
 - Prevented workflow stage sessions from exposing or executing the `workflow` tool while preserving stage-level subagent delegation.
-- Retained completed, failed, and killed workflow runs in user-facing status/connect surfaces and changed workflow kill controls to mark runs killed without removing them from live inspection history ([#1083](https://github.com/flora131/atomic/issues/1083)).
+- Retained completed, failed, and killed workflow runs in user-facing status/connect surfaces and changed workflow kill controls to mark runs killed without removing them from live inspection history ([#1083](https://github.com/bastani-inc/atomic/issues/1083)).
 
 ## [0.8.18] - 2026-05-27
 
@@ -85,11 +151,11 @@ The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.1.0/),
 
 ### Breaking Changes
 
-- Removed Ralph's configurable `review_quorum` and `blocker_threshold` inputs; `max_turns` remains configurable while reviewer quorum and blocker threshold use fixed controller defaults ([#1061](https://github.com/flora131/atomic/issues/1061)).
+- Removed Ralph's configurable `review_quorum` and `blocker_threshold` inputs; `max_turns` remains configurable while reviewer quorum and blocker threshold use fixed controller defaults ([#1061](https://github.com/bastani-inc/atomic/issues/1061)).
 
 ### Changed
 
-- Aligned Ralph's goal-continuation prompt more closely with Codex `/goal` hidden continuation guidance, including current-state evidence, non-shrinking scope, fidelity, completion audit, and strict blocked audit language ([#1061](https://github.com/flora131/atomic/issues/1061)).
+- Aligned Ralph's goal-continuation prompt more closely with Codex `/goal` hidden continuation guidance, including current-state evidence, non-shrinking scope, fidelity, completion audit, and strict blocked audit language ([#1061](https://github.com/bastani-inc/atomic/issues/1061)).
 - Restored Ralph's stronger historical review gate prompt and `review_decision` schema with findings, oracle satisfaction, receipt assessment, verification remaining, and reviewer-error guard fields.
 
 ## [0.8.16] - 2026-05-26
@@ -102,47 +168,47 @@ The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.1.0/),
 
 ### Breaking Changes
 
-- Removed Ralph's `prompt` and `max_loops` compatibility inputs; use the canonical `objective` and `max_turns` inputs instead ([#1061](https://github.com/flora131/atomic/issues/1061)).
+- Removed Ralph's `prompt` and `max_loops` compatibility inputs; use the canonical `objective` and `max_turns` inputs instead ([#1061](https://github.com/bastani-inc/atomic/issues/1061)).
 
 ### Changed
 
-- Reworked Ralph into a Goal Runner workflow with a persisted goal ledger, bounded worker turns, parallel structured reviewers, reviewer-quorum completion, repeated-blocker gating, and final status reporting while excluding token-budget behavior ([#1061](https://github.com/flora131/atomic/issues/1061)).
+- Reworked Ralph into a Goal Runner workflow with a persisted goal ledger, bounded worker turns, parallel structured reviewers, reviewer-quorum completion, repeated-blocker gating, and final status reporting while excluding token-budget behavior ([#1061](https://github.com/bastani-inc/atomic/issues/1061)).
 
 ### Fixed
 
-- Hardened Ralph's goal-runner edge cases around base-branch sanitization coverage, worker/reviewer failure handling, cross-turn review context, bounded blocker-threshold handling, and unused ledger state ([#1061](https://github.com/flora131/atomic/issues/1061)).
+- Hardened Ralph's goal-runner edge cases around base-branch sanitization coverage, worker/reviewer failure handling, cross-turn review context, bounded blocker-threshold handling, and unused ledger state ([#1061](https://github.com/bastani-inc/atomic/issues/1061)).
 
 ## [0.8.15] - 2026-05-26
 
 ### Breaking Changes
 
-- Changed workflow human-in-the-loop prompts (`ctx.ui.input`, `ctx.ui.confirm`, `ctx.ui.select`, and `ctx.ui.editor`) to render as synthetic graph stages with `awaiting_input` status; consumers should handle the new `StageSnapshot.promptAnswerState` metadata and avoid caching `StageSnapshot.parentIds` references across store updates ([#1046](https://github.com/flora131/atomic/issues/1046)).
-- Changed `ctx.ui.select(..., [])` to throw `pi-workflows: ctx.ui.select requires at least one option` before creating a prompt node instead of returning an empty string ([#1046](https://github.com/flora131/atomic/issues/1046)).
+- Changed workflow human-in-the-loop prompts (`ctx.ui.input`, `ctx.ui.confirm`, `ctx.ui.select`, and `ctx.ui.editor`) to render as synthetic graph stages with `awaiting_input` status; consumers should handle the new `StageSnapshot.promptAnswerState` metadata and avoid caching `StageSnapshot.parentIds` references across store updates ([#1046](https://github.com/bastani-inc/atomic/issues/1046)).
+- Changed `ctx.ui.select(..., [])` to throw `pi-workflows: ctx.ui.select requires at least one option` before creating a prompt node instead of returning an empty string ([#1046](https://github.com/bastani-inc/atomic/issues/1046)).
 
 ### Added
 
-- Added workflow tool stage introspection and control actions (`stages`, `stage`, `transcript`, `send`, `pause`, and `reload`) for inspecting stage state, reading transcripts, answering prompts, controlling live stages, pausing runs, and reloading workflow resources ([#1023](https://github.com/flora131/atomic/issues/1023)).
-- Added live-memory prompt answer replay for workflow continuations using snapshot-safe `promptAnswerState` markers and a private `PromptAnswerRecord` ledger that is excluded from snapshots and persistence ([#1046](https://github.com/flora131/atomic/issues/1046)).
+- Added workflow tool stage introspection and control actions (`stages`, `stage`, `transcript`, `send`, `pause`, and `reload`) for inspecting stage state, reading transcripts, answering prompts, controlling live stages, pausing runs, and reloading workflow resources ([#1023](https://github.com/bastani-inc/atomic/issues/1023)).
+- Added live-memory prompt answer replay for workflow continuations using snapshot-safe `promptAnswerState` markers and a private `PromptAnswerRecord` ledger that is excluded from snapshots and persistence ([#1046](https://github.com/bastani-inc/atomic/issues/1046)).
 
 ### Changed
 
-- Updated Ralph around an autonomous goal contract so planner, implementation, and reviewer/auditor stages infer verifiable criteria, require receipts, and judge completion against the verification oracle instead of document completeness alone ([#1053](https://github.com/flora131/atomic/issues/1053)).
-- Updated Ralph's orchestrator prompt to discover project initialization needs from repository evidence before implementation work proceeds ([#1048](https://github.com/flora131/atomic/issues/1048)).
+- Updated Ralph around an autonomous goal contract so planner, implementation, and reviewer/auditor stages infer verifiable criteria, require receipts, and judge completion against the verification oracle instead of document completeness alone ([#1053](https://github.com/bastani-inc/atomic/issues/1053)).
+- Updated Ralph's orchestrator prompt to discover project initialization needs from repository evidence before implementation work proceeds ([#1048](https://github.com/bastani-inc/atomic/issues/1048)).
 - Added a final Ralph PR-preparation phase that reviews changes against the configured base branch, tries available GitHub credentials using local git identity as a hint, posts implementation notes as a PR comment, and creates a pull request when possible.
-- Updated Ralph to persist planner goal contracts and implementation notes as OS-temp workflow artifacts, and pass file paths rather than full plan text to later stages ([#1037](https://github.com/flora131/atomic/issues/1037)).
+- Updated Ralph to persist planner goal contracts and implementation notes as OS-temp workflow artifacts, and pass file paths rather than full plan text to later stages ([#1037](https://github.com/bastani-inc/atomic/issues/1037)).
 - Updated `deep-research-codebase` output layout to write public reports under `research/` and hidden per-run handoff artifacts under `research/.deep-research-<run-id>/`.
 
 ### Fixed
 
 - Returned the workflow overlay to the graph orchestrator after answering or skipping a `ctx.ui` prompt node, and preserved the prompt's question footprint for read-only prompt-node views.
-- Removed biasing stage-output and iteration-count context from Ralph reviewer prompts while making the comparison base branch explicit ([#1037](https://github.com/flora131/atomic/issues/1037)).
-- Ordered snapshot transcript fallback entries chronologically before applying `tail`/`limit`, preserving terminal result/error entries after tools for missing or tied timestamps ([#1023](https://github.com/flora131/atomic/issues/1023)).
-- Reloaded workflow resources directly for `workflow({ action: "reload" })` instead of queuing a literal `/workflow reload` follow-up ([#1023](https://github.com/flora131/atomic/issues/1023)).
-- Kept pending prompts unresolved when `workflow({ action: "send" })` omits `text`, `response`, and `message`, while preserving explicit empty-string answers ([#1023](https://github.com/flora131/atomic/issues/1023)).
+- Removed biasing stage-output and iteration-count context from Ralph reviewer prompts while making the comparison base branch explicit ([#1037](https://github.com/bastani-inc/atomic/issues/1037)).
+- Ordered snapshot transcript fallback entries chronologically before applying `tail`/`limit`, preserving terminal result/error entries after tools for missing or tied timestamps ([#1023](https://github.com/bastani-inc/atomic/issues/1023)).
+- Reloaded workflow resources directly for `workflow({ action: "reload" })` instead of queuing a literal `/workflow reload` follow-up ([#1023](https://github.com/bastani-inc/atomic/issues/1023)).
+- Kept pending prompts unresolved when `workflow({ action: "send" })` omits `text`, `response`, and `message`, while preserving explicit empty-string answers ([#1023](https://github.com/bastani-inc/atomic/issues/1023)).
 - Included `deep-research-codebase` discovery-stage handoff files in the persisted run manifest.
 - Persisted `deep-research-codebase` final reports as dated Markdown research docs while retaining file-only handoffs for bounded aggregation.
-- Prevented `deep-research-codebase` aggregation from inlining large specialist transcripts by using file-only handoff artifacts ([#1016](https://github.com/flora131/atomic/issues/1016)).
-- Removed model metadata from workflow node cards while retaining fallback dependency metadata ([#1011](https://github.com/flora131/atomic/issues/1011)).
+- Prevented `deep-research-codebase` aggregation from inlining large specialist transcripts by using file-only handoff artifacts ([#1016](https://github.com/bastani-inc/atomic/issues/1016)).
+- Removed model metadata from workflow node cards while retaining fallback dependency metadata ([#1011](https://github.com/bastani-inc/atomic/issues/1011)).
 - Preserve the selected workflow switcher row highlight through truncation ellipses on long stage names.
 
 ## [0.0.1] — 2026-05-15