@mediadatafusion/pi-workflow-suite 0.0.14 → 0.0.16

package/CHANGELOG.md

@@ -2,6 +2,32 @@
 
 All notable public releases will be documented in this file.
 
+## [0.0.16] - 2026-06-09
+
+### Fixed
+
+- Fixed Plan Mode recovery so interrupted or resumed workflows return to the correct state more reliably.
+- Fixed Standard Mode completion cleanup to prevent stale active-status indicators after work is complete.
+- Fixed workflow preset behavior so presets preserve user-selected model-routing preferences.
+- Improved compaction recovery for continued workflow sessions.
+- Refined repair flow handling so advisory follow-up notes do not unnecessarily block continued progress.
+
+### Improved
+
+- Improved overall workflow reliability across Plan, Standard, and repair flows.
+
+## [0.0.15] - 2026-06-09
+
+### Improved
+
+- Improved Plan Mode approval handoffs so approval-ready plans remain visible while action menus are open.
+- Refined the Deep Plan preset so reviewer review remains available without automatically starting by default.
+- Improved workflow recovery around transient connection interruptions so Plan, Mission, and Standard workflows are easier to resume after short transport failures.
+
+### Hardened
+
+- Expanded regression coverage for reviewer routing, sub-agent handoffs, package command surfaces, and interruption recovery.
+
 ## [0.0.14] - 2026-06-08
 
 ### Fixed

package/README.md

@@ -4,7 +4,7 @@
 
 [![Install](https://cdn.jsdelivr.net/npm/@mediadatafusion/pi-workflow-suite@0.0.12/docs/assets/readme-link-install.svg)](#installation) [![Quick Start](https://cdn.jsdelivr.net/npm/@mediadatafusion/pi-workflow-suite@0.0.12/docs/assets/readme-link-quick-start.svg)](#quick-start) [![Commands](https://cdn.jsdelivr.net/npm/@mediadatafusion/pi-workflow-suite@0.0.12/docs/assets/readme-link-commands.svg)](#core-commands) [![Settings](https://cdn.jsdelivr.net/npm/@mediadatafusion/pi-workflow-suite@0.0.12/docs/assets/readme-link-settings.svg)](#settings-reference)
 
-**Workflow Suite Version:** `v0.0.14`
+**Workflow Suite Version:** `v0.0.16`
 
 ## Overview
 
@@ -578,15 +578,15 @@ Mission: approval-gated, auto-run after approval, standard planning, milestone v
 
 Deep — Careful
 Plan: deep, asks clarification for non-trivial work
-Review: automatic
-Validation: automatic
+Review: manual/optional; not automatic before execution
+Validation: automatic after execution
 Sub-agents: forced larger teams across planning/execution/repair/review/validation
 Mission: deep planning with final validation
 
 Maximum — Thorough
 Plan: maximum
 Review: automatic
-Validation: automatic
+Validation: automatic after execution
 Sub-agents: forced maximum teams across planning/execution/repair/review/validation
 Mission: supervised auto, final validation, higher retry budget
 ```
@@ -1007,8 +1007,8 @@ pi install -l npm:@mediadatafusion/pi-workflow-suite
 ### Installing specific versions
 
 ```bash
-pi install npm:@mediadatafusion/pi-workflow-suite@0.0.14
-pi install -l npm:@mediadatafusion/pi-workflow-suite@0.0.14
+pi install npm:@mediadatafusion/pi-workflow-suite@0.0.16
+pi install -l npm:@mediadatafusion/pi-workflow-suite@0.0.16
 ```
 
 An unversioned install follows normal update behavior: `pi update` and `pi update --extensions` will pick up new package releases. A versioned install pins the package to that version. Pinned package specs are intentionally skipped by Pi's normal package update commands. To move a pinned install to a newer version, reinstall with the desired version. To switch back to latest tracking, use the unversioned install command without `@<version>`.
@@ -1214,10 +1214,10 @@ See `docs/TROUBLESHOOTING.md` for detailed diagnostics.
 
 ## Versioning
 
-The current preparation version is `v0.0.14`. Version information is intentionally aligned across:
+The current preparation version is `v0.0.16`. Version information is intentionally aligned across:
 
-- `VERSION` (`v0.0.14`),
-- `package.json` (`0.0.14`),
+- `VERSION` (`v0.0.16`),
+- `package.json` (`0.0.16`),
 - `package-lock.json`,
 - this README,
 - Workflow Suite settings/about output.

package/VERSION

@@ -1 +1 @@
-v0.0.14
+v0.0.16

package/config/prompts/mission-repair.md

@@ -4,7 +4,7 @@ MANDATORY STRUCTURED HANDOFF: call workflow_repair_result before final response
 
 You are PI MISSION MODE REPAIR EXECUTOR.
 
-Repair only concrete validator-identified failures for the current mission milestone. Do not re-grade validation; only Mission validation can pass repaired work.
+Repair concrete validator-identified failures for the current mission milestone and keep going while fixes are in-scope and non-destructive. Do not re-grade validation; only Mission validation can pass repaired work.
 
 Rules:
 - Only fix concrete issues directly related to the failed milestone validation.
@@ -32,6 +32,8 @@ Use only these exact installed agent names when calling the subagent tool. Do no
 - Do not create arbitrary repository-root files. A root file is allowed only when the mission plan, user request, or validator finding names that exact root path.
 - If a current-task-created file is in the wrong location but contains recoverable work, move or rename it to the correct approved location instead of deleting it.
 - Treat untracked, unexpected, or ambiguous files as possibly user-owned; do not delete, overwrite, move, or clean them without explicit approval for that exact file.
+- Do not set `needsUserApproval` for advisory-only follow-up, credential rotation recommendations, preserved ambiguous files, manual QA still needed, or pre-existing project debt. Put those in the summary/safety notes and let revalidation run.
+- Set `needsUserApproval` only when a concrete hard-safety action or artifact disposition should pause automatic revalidation.
 - Keep file writes sequential unless workflow settings explicitly allow safe scoped parallel edits.
 - Preserve checkpoint integrity and disclose moved, preserved, deleted, root, and possibly user-owned files.
 

package/config/prompts/workflow-repair.md

@@ -6,7 +6,7 @@ You are PI WORKFLOW REPAIR MODE.
 
 Available tools in repair mode: edit, write, bash, workflow_diagram, workflow_progress, workflow_repair_result. The workflow_repair_result tool IS registered and active. If you cannot see it in your tool list, re-check — it is available. You MUST call it with your repair summary before finishing. Do not output a prose-only repair report; use the typed handoff tool.
 
-Repair only concrete validator-identified failed validation items for the approved Plan Mode workflow. Do not re-grade validation; only the validator/revalidator can declare PASS.
+Repair concrete validator-identified failed validation items for the approved Plan Mode workflow and keep going while fixes are in-scope and non-destructive. Do not re-grade validation; only the validator/revalidator can declare PASS.
 
 Rules:
 - The approved plan is still the execution contract.
@@ -14,11 +14,13 @@ Rules:
 - Do not perform unrelated refactors.
 - Do not commit, push, deploy, or mutate databases.
 - Do not edit secrets, auth/session files, runtime logs/state, `.env`, `.factory`, or `.cursor` files.
-- Stop and report if the repair requires destructive, out-of-scope, secret-adjacent, deployment, database, or otherwise risky action.
+- Stop and report if the repair requires destructive, out-of-scope, secret-adjacent, deployment, database, protected runtime/auth/session, or otherwise truly risky action.
 - Do not create arbitrary repository-root files. A root file is allowed only when the approved plan, user request, or validator finding names that exact root path.
 - If a current-task-created file is in the wrong location but contains recoverable work, move or rename it to the correct approved location instead of deleting it.
 - Treat untracked, unexpected, or ambiguous files as possibly user-owned; do not delete, overwrite, move, or clean them without explicit approval for that exact file.
 - If the validation finding is only manual/visual/browser verification or says no code repair is needed, do not change code; summarize manual QA/revalidation readiness.
+- Do not set `needsUserApproval` for advisory-only follow-up, credential rotation recommendations, preserved ambiguous files, manual QA still needed, or pre-existing project debt. Put those in the summary/safety notes and let revalidation run.
+- Set `needsUserApproval` only when a concrete hard-safety action or artifact disposition should pause automatic revalidation.
 - Use repair sub-agents aggressively for failure triage, missing-file inspection, patch planning, and validation preparation when policy allows/requires them.
 
 ## Available Sub-Agent Types

package/extensions/workflow-model-router.ts

@@ -280,8 +280,8 @@ export interface WorkflowPresetBundle {
   description?: string;
   planning?: Partial<WorkflowSettings["planning"]>;
   workflow?: Partial<WorkflowSettings["workflow"]>;
-  standard?: Partial<Omit<WorkflowSettings["standard"], "models">> & { models?: Partial<Record<WorkflowRole, RoleModelSettings>> };
-  missions?: Partial<Omit<WorkflowSettings["missions"], "models">>;
+  standard?: Partial<Omit<WorkflowSettings["standard"], "models" | "useSharedExecutorModel" | "useStandardSpecificModels" | "modelRole">>;
+  missions?: Partial<Omit<WorkflowSettings["missions"], "models" | "useMissionSpecificModels">>;
   subagents?: Partial<WorkflowSettings["subagents"]>;
   ui?: Partial<WorkflowSettings["ui"]>;
 }
@@ -801,29 +801,55 @@ interface UpdateSettingsOptions {
   preserveActivePreset?: boolean;
 }
 
-export const WORKFLOW_MANUAL_PRESET = "custom";
+export const WORKFLOW_CUSTOM_PRESET_MARKER = "custom";
+
+function standardSettingsWithoutModelRouting(standard?: Partial<WorkflowSettings["standard"]>): Partial<WorkflowSettings["standard"]> {
+  const {
+    models: _ignoredModels,
+    useSharedExecutorModel: _ignoredUseSharedExecutorModel,
+    useStandardSpecificModels: _ignoredUseStandardSpecificModels,
+    modelRole: _ignoredModelRole,
+    ...standardWithoutModelRouting
+  } = standard ?? {};
+  return standardWithoutModelRouting;
+}
+
+function missionSettingsWithoutModelRouting(missions?: Partial<WorkflowSettings["missions"]>): Partial<WorkflowSettings["missions"]> {
+  const {
+    models: _ignoredModels,
+    useMissionSpecificModels: _ignoredUseMissionSpecificModels,
+    ...missionsWithoutModelRouting
+  } = missions ?? {};
+  return missionsWithoutModelRouting;
+}
 
 function presetOwnedSettingsSignature(settings: WorkflowSettings): string {
-  const { models: _ignoredStandardModels, ...standardWithoutModels } = settings.standard;
-  const { models: _ignoredMissionModels, ...missionsWithoutModels } = settings.missions;
   return JSON.stringify({
     planning: settings.planning,
     workflow: settings.workflow,
-    standard: standardWithoutModels,
-    missions: missionsWithoutModels,
+    standard: standardSettingsWithoutModelRouting(settings.standard),
+    missions: missionSettingsWithoutModelRouting(settings.missions),
     subagents: settings.subagents,
   });
 }
 
 function applyActivePresetOverlay(settings: WorkflowSettings): WorkflowSettings {
-  const active = settings.presets?.activePreset ?? WORKFLOW_MANUAL_PRESET;
-  if (active === WORKFLOW_MANUAL_PRESET) return settings;
+  const active = settings.presets?.activePreset ?? WORKFLOW_CUSTOM_PRESET_MARKER;
+  if (active === WORKFLOW_CUSTOM_PRESET_MARKER) return settings;
   const preset = workflowPresetCatalog(settings)[active];
   if (!preset) return settings;
   const defaults = defaultWorkflowSettings();
   const normalized = normalizeWorkflowPresetBundle(preset);
-  const currentStandardModels = settings.standard.models;
-  const currentMissionModels = settings.missions.models;
+  const currentStandardModelRouting = {
+    models: settings.standard.models,
+    useSharedExecutorModel: settings.standard.useSharedExecutorModel,
+    useStandardSpecificModels: settings.standard.useStandardSpecificModels,
+    modelRole: settings.standard.modelRole,
+  };
+  const currentMissionModelRouting = {
+    models: settings.missions.models,
+    useMissionSpecificModels: settings.missions.useMissionSpecificModels,
+  };
   return {
     ...settings,
     planning: normalized.planning ? { ...settings.planning, ...normalized.planning } : settings.planning,
@@ -833,9 +859,9 @@ function applyActivePresetOverlay(settings: WorkflowSettings): WorkflowSettings
       ...settings.standard,
       ...normalized.standard,
       subagents: { ...(defaults.standard.subagents ?? {}), ...(settings.standard.subagents ?? {}), ...(normalized.standard.subagents ?? {}) },
-      models: currentStandardModels,
+      ...currentStandardModelRouting,
     },
-    missions: normalized.missions ? { ...settings.missions, ...normalized.missions, models: currentMissionModels } : settings.missions,
+    missions: normalized.missions ? { ...settings.missions, ...normalized.missions, ...currentMissionModelRouting } : settings.missions,
     subagents: normalized.subagents ? { ...settings.subagents, ...normalized.subagents } : settings.subagents,
     presets: settings.presets,
   };
@@ -923,12 +949,12 @@ export function createProjectSettingsOverride(cwd: string): SettingsWriteResult
 export function updateSettings(cwd: string, requestedScope: WorkflowSettingsScope | undefined, updater: (settings: WorkflowSettings) => void, options: UpdateSettingsOptions = {}): SettingsWriteResult {
   const target = getWriteTarget(cwd, requestedScope);
   const settings = target.scope === "global" ? loadGlobalSettings() : loadEffectiveSettings(cwd).settings;
-  const beforePreset = settings.presets?.activePreset ?? WORKFLOW_MANUAL_PRESET;
+  const beforePreset = settings.presets?.activePreset ?? WORKFLOW_CUSTOM_PRESET_MARKER;
   const beforePresetOwned = presetOwnedSettingsSignature(settings);
   updater(settings);
   const afterPresetOwned = presetOwnedSettingsSignature(settings);
-  if (!options.preserveActivePreset && beforePreset !== WORKFLOW_MANUAL_PRESET && settings.presets?.activePreset === beforePreset && beforePresetOwned !== afterPresetOwned) {
-    settings.presets = { ...(settings.presets ?? {}), activePreset: WORKFLOW_MANUAL_PRESET, items: { ...(settings.presets?.items ?? {}) } };
+  if (!options.preserveActivePreset && beforePreset !== WORKFLOW_CUSTOM_PRESET_MARKER && settings.presets?.activePreset === beforePreset && beforePresetOwned !== afterPresetOwned) {
+    settings.presets = { ...(settings.presets ?? {}), activePreset: WORKFLOW_CUSTOM_PRESET_MARKER, items: { ...(settings.presets?.items ?? {}) } };
   }
   saveSettingsFile(target.file, settings);
   return { settings, scope: target.scope, file: target.file };
@@ -941,7 +967,7 @@ export function builtInWorkflowPresets(): Record<string, WorkflowPresetBundle> {
       description: "Fast end-to-end Plan/Mission/Standard workflow with minimal ceremony, automatic validation when work runs, low safe repair retries, and one-worker sub-agent support.",
       planning: { depth: "fast", clarificationMode: "auto", maxClarificationQuestions: 2, interactiveClarificationEnabled: true, clarificationQualityGate: false, useSubagentsBeforeClarification: true },
       workflow: { offerReviewerBeforeExecute: false, autoRunReviewerBeforeExecute: false, offerValidationAfterExecute: true, autoRunValidationAfterExecute: true, validateAfterExecution: true, requirePlanApprovalBeforeExecute: false, requireApprovalBeforeExecution: false, autoRepairReviewFailures: false, autoRepairValidationFailures: true, reviewRetryMode: "off", validationRetryMode: "safe_only", maxReviewRetriesPerPlan: 0, maxReviewRetriesPerWorkflow: 0, maxValidationRetriesPerPlan: 1, maxValidationRetriesPerWorkflow: 2, pauseAfterReviewFailure: true, pauseAfterValidationFailure: false, planProgressEnabled: true, planRuntimeEnabled: true, planShowProgressBar: true },
-      standard: { enabled: true, autoTodoEnabled: true, todoProgressVisible: true, todoTriggerMode: "auto", clarificationEnabled: true, clarificationMode: "auto", maxClarificationQuestions: 1, interactiveClarificationEnabled: true, clarificationTiming: "after_initial_analysis", clarificationQualityGate: false, allowClarificationWithoutAnalysis: false, useSubagentsBeforeClarification: false, allowSubagents: true, subagentScope: "user", subagents: { planningPolicy: "forced", executionPolicy: "forced", repairPolicy: "forced", reviewPolicy: "auto", validationPolicy: "forced", autoUseDuringPlanning: true, autoUseDuringExecution: true, autoUseDuringRepair: true, autoUseDuringReview: true, autoUseDuringValidation: true, minPlanningWorkersForDeep: 1, minPlanningWorkersForMaximum: 1, minExecutionWorkersForDeep: 1, minExecutionWorkersForMaximum: 1, minRepairWorkersForDeep: 1, minRepairWorkersForMaximum: 1, minReviewWorkersForDeep: 1, minReviewWorkersForMaximum: 1, minValidationWorkersForDeep: 1, minValidationWorkersForMaximum: 1 }, statusWidgetVisible: true, useSharedExecutorModel: true, useStandardSpecificModels: false, modelRole: "executor" },
+      standard: { enabled: true, autoTodoEnabled: true, todoProgressVisible: true, todoTriggerMode: "auto", clarificationEnabled: true, clarificationMode: "auto", maxClarificationQuestions: 1, interactiveClarificationEnabled: true, clarificationTiming: "after_initial_analysis", clarificationQualityGate: false, allowClarificationWithoutAnalysis: false, useSubagentsBeforeClarification: false, allowSubagents: true, subagentScope: "user", subagents: { planningPolicy: "forced", executionPolicy: "forced", repairPolicy: "forced", reviewPolicy: "auto", validationPolicy: "forced", autoUseDuringPlanning: true, autoUseDuringExecution: true, autoUseDuringRepair: true, autoUseDuringReview: true, autoUseDuringValidation: true, minPlanningWorkersForDeep: 1, minPlanningWorkersForMaximum: 1, minExecutionWorkersForDeep: 1, minExecutionWorkersForMaximum: 1, minRepairWorkersForDeep: 1, minRepairWorkersForMaximum: 1, minReviewWorkersForDeep: 1, minReviewWorkersForMaximum: 1, minValidationWorkersForDeep: 1, minValidationWorkersForMaximum: 1 }, statusWidgetVisible: true },
       missions: { defaultAutonomy: "approval_gated", requireValidationPerMilestone: true, autoRunAfterApproval: true, continueAcrossMilestones: true, pauseBetweenMilestones: false, clarificationMode: "auto", maxClarificationQuestions: 2, planningDepth: "fast", useSubagentsBeforeClarification: true, autoRepairReviewFailures: false, reviewRetryMode: "off", maxReviewRetriesPerMission: 0, autoRepairValidationFailures: true, validationRetryMode: "safe_only", maxValidationRetriesPerMilestone: 1, maxValidationRetriesPerMission: 2, finalValidationEnabled: false, autoRepairFinalValidationFailures: false, maxFinalValidationRetries: 0, subagentPolicy: "forced", minWorkersForDeep: 1, minWorkersForMaximum: 1 },
       subagents: { planningPolicy: "forced", executionPolicy: "forced", repairPolicy: "forced", reviewPolicy: "auto", validationPolicy: "forced", autoUseDuringPlanning: true, autoUseDuringExecution: true, autoUseDuringRepair: true, autoUseDuringReview: true, autoUseDuringValidation: true, minPlanningWorkersForDeep: 1, minPlanningWorkersForMaximum: 1, minExecutionWorkersForDeep: 1, minExecutionWorkersForMaximum: 1, minRepairWorkersForDeep: 1, minRepairWorkersForMaximum: 1, minReviewWorkersForDeep: 1, minReviewWorkersForMaximum: 1, minValidationWorkersForDeep: 1, minValidationWorkersForMaximum: 1, allowBackgroundSubagents: false },
     },
@@ -950,16 +976,16 @@ export function builtInWorkflowPresets(): Record<string, WorkflowPresetBundle> {
       description: "Default end-to-end workflow with useful clarification, automatic execution/validation after approval, safe repair retries, and balanced worker support.",
       planning: { depth: "standard", clarificationMode: "auto", maxClarificationQuestions: 3, interactiveClarificationEnabled: true, clarificationQualityGate: true, useSubagentsBeforeClarification: true },
       workflow: { offerReviewerBeforeExecute: false, autoRunReviewerBeforeExecute: false, offerValidationAfterExecute: true, autoRunValidationAfterExecute: true, validateAfterExecution: true, requirePlanApprovalBeforeExecute: false, requireApprovalBeforeExecution: false, autoRepairReviewFailures: true, autoRepairValidationFailures: true, reviewRetryMode: "safe_only", validationRetryMode: "safe_only", maxReviewRetriesPerPlan: 2, maxReviewRetriesPerWorkflow: 4, maxValidationRetriesPerPlan: 2, maxValidationRetriesPerWorkflow: 4, pauseAfterReviewFailure: false, pauseAfterValidationFailure: false, planProgressEnabled: true, planRuntimeEnabled: true, planShowProgressBar: true },
-      standard: { enabled: true, autoTodoEnabled: true, todoProgressVisible: true, todoTriggerMode: "auto", clarificationEnabled: true, clarificationMode: "auto", maxClarificationQuestions: 1, interactiveClarificationEnabled: true, clarificationTiming: "after_initial_analysis", clarificationQualityGate: true, allowClarificationWithoutAnalysis: false, useSubagentsBeforeClarification: false, allowSubagents: true, subagentScope: "user", subagents: { planningPolicy: "forced", executionPolicy: "forced", repairPolicy: "forced", reviewPolicy: "forced", validationPolicy: "forced", autoUseDuringPlanning: true, autoUseDuringExecution: true, autoUseDuringRepair: true, autoUseDuringReview: true, autoUseDuringValidation: true, minPlanningWorkersForDeep: 1, minPlanningWorkersForMaximum: 1, minExecutionWorkersForDeep: 2, minExecutionWorkersForMaximum: 2, minRepairWorkersForDeep: 2, minRepairWorkersForMaximum: 2, minReviewWorkersForDeep: 2, minReviewWorkersForMaximum: 2, minValidationWorkersForDeep: 2, minValidationWorkersForMaximum: 2 }, statusWidgetVisible: true, useSharedExecutorModel: true, useStandardSpecificModels: false, modelRole: "executor" },
+      standard: { enabled: true, autoTodoEnabled: true, todoProgressVisible: true, todoTriggerMode: "auto", clarificationEnabled: true, clarificationMode: "auto", maxClarificationQuestions: 1, interactiveClarificationEnabled: true, clarificationTiming: "after_initial_analysis", clarificationQualityGate: true, allowClarificationWithoutAnalysis: false, useSubagentsBeforeClarification: false, allowSubagents: true, subagentScope: "user", subagents: { planningPolicy: "forced", executionPolicy: "forced", repairPolicy: "forced", reviewPolicy: "forced", validationPolicy: "forced", autoUseDuringPlanning: true, autoUseDuringExecution: true, autoUseDuringRepair: true, autoUseDuringReview: true, autoUseDuringValidation: true, minPlanningWorkersForDeep: 1, minPlanningWorkersForMaximum: 1, minExecutionWorkersForDeep: 2, minExecutionWorkersForMaximum: 2, minRepairWorkersForDeep: 2, minRepairWorkersForMaximum: 2, minReviewWorkersForDeep: 2, minReviewWorkersForMaximum: 2, minValidationWorkersForDeep: 2, minValidationWorkersForMaximum: 2 }, statusWidgetVisible: true },
       missions: { defaultAutonomy: "approval_gated", requireValidationPerMilestone: true, autoRunAfterApproval: true, continueAcrossMilestones: true, pauseBetweenMilestones: false, clarificationMode: "auto", maxClarificationQuestions: 3, planningDepth: "standard", useSubagentsBeforeClarification: true, autoRepairReviewFailures: true, reviewRetryMode: "safe_only", maxReviewRetriesPerMission: 2, autoRepairValidationFailures: true, validationRetryMode: "safe_only", maxValidationRetriesPerMilestone: 2, maxValidationRetriesPerMission: 6, finalValidationEnabled: false, autoRepairFinalValidationFailures: false, maxFinalValidationRetries: 1, subagentPolicy: "forced", minWorkersForDeep: 1, minWorkersForMaximum: 1 },
       subagents: { planningPolicy: "forced", executionPolicy: "forced", repairPolicy: "forced", reviewPolicy: "forced", validationPolicy: "forced", autoUseDuringPlanning: true, autoUseDuringExecution: true, autoUseDuringRepair: true, autoUseDuringReview: true, autoUseDuringValidation: true, minPlanningWorkersForDeep: 1, minPlanningWorkersForMaximum: 1, minExecutionWorkersForDeep: 2, minExecutionWorkersForMaximum: 2, minRepairWorkersForDeep: 2, minRepairWorkersForMaximum: 2, minReviewWorkersForDeep: 2, minReviewWorkersForMaximum: 2, minValidationWorkersForDeep: 2, minValidationWorkersForMaximum: 2, allowBackgroundSubagents: false },
     },
     deep: {
       displayName: "Deep",
-      description: "Careful end-to-end workflow for risky or codebase-heavy work with stronger clarification, automatic review/validation, final mission validation, and larger worker teams.",
+      description: "Careful end-to-end workflow for risky or codebase-heavy work with stronger clarification, manual Plan review, automatic validation, final mission validation, and larger worker teams.",
       planning: { depth: "deep", clarificationMode: "always_for_nontrivial", maxClarificationQuestions: 5, interactiveClarificationEnabled: true, clarificationQualityGate: true, useSubagentsBeforeClarification: true },
-      workflow: { offerReviewerBeforeExecute: false, autoRunReviewerBeforeExecute: true, offerValidationAfterExecute: true, autoRunValidationAfterExecute: true, validateAfterExecution: true, requirePlanApprovalBeforeExecute: false, requireApprovalBeforeExecution: false, autoRepairReviewFailures: true, autoRepairValidationFailures: true, reviewRetryMode: "safe_only", validationRetryMode: "safe_only", maxReviewRetriesPerPlan: 3, maxReviewRetriesPerWorkflow: 6, maxValidationRetriesPerPlan: 3, maxValidationRetriesPerWorkflow: 6, pauseAfterReviewFailure: false, pauseAfterValidationFailure: false, planProgressEnabled: true, planRuntimeEnabled: true, planShowProgressBar: true },
-      standard: { enabled: true, autoTodoEnabled: true, todoProgressVisible: true, todoTriggerMode: "auto", clarificationEnabled: true, clarificationMode: "auto", maxClarificationQuestions: 2, interactiveClarificationEnabled: true, clarificationTiming: "after_initial_analysis", clarificationQualityGate: true, allowClarificationWithoutAnalysis: false, useSubagentsBeforeClarification: true, allowSubagents: true, subagentScope: "user", subagents: { planningPolicy: "forced", executionPolicy: "forced", repairPolicy: "forced", reviewPolicy: "forced", validationPolicy: "forced", autoUseDuringPlanning: true, autoUseDuringExecution: true, autoUseDuringRepair: true, autoUseDuringReview: true, autoUseDuringValidation: true, minPlanningWorkersForDeep: 2, minPlanningWorkersForMaximum: 2, minExecutionWorkersForDeep: 3, minExecutionWorkersForMaximum: 3, minRepairWorkersForDeep: 2, minRepairWorkersForMaximum: 2, minReviewWorkersForDeep: 3, minReviewWorkersForMaximum: 3, minValidationWorkersForDeep: 3, minValidationWorkersForMaximum: 3 }, statusWidgetVisible: true, useSharedExecutorModel: true, useStandardSpecificModels: false, modelRole: "executor" },
+      workflow: { offerReviewerBeforeExecute: false, autoRunReviewerBeforeExecute: false, offerValidationAfterExecute: true, autoRunValidationAfterExecute: true, validateAfterExecution: true, requirePlanApprovalBeforeExecute: false, requireApprovalBeforeExecution: false, autoRepairReviewFailures: true, autoRepairValidationFailures: true, reviewRetryMode: "safe_only", validationRetryMode: "safe_only", maxReviewRetriesPerPlan: 3, maxReviewRetriesPerWorkflow: 6, maxValidationRetriesPerPlan: 3, maxValidationRetriesPerWorkflow: 6, pauseAfterReviewFailure: false, pauseAfterValidationFailure: false, planProgressEnabled: true, planRuntimeEnabled: true, planShowProgressBar: true },
+      standard: { enabled: true, autoTodoEnabled: true, todoProgressVisible: true, todoTriggerMode: "auto", clarificationEnabled: true, clarificationMode: "auto", maxClarificationQuestions: 2, interactiveClarificationEnabled: true, clarificationTiming: "after_initial_analysis", clarificationQualityGate: true, allowClarificationWithoutAnalysis: false, useSubagentsBeforeClarification: true, allowSubagents: true, subagentScope: "user", subagents: { planningPolicy: "forced", executionPolicy: "forced", repairPolicy: "forced", reviewPolicy: "forced", validationPolicy: "forced", autoUseDuringPlanning: true, autoUseDuringExecution: true, autoUseDuringRepair: true, autoUseDuringReview: true, autoUseDuringValidation: true, minPlanningWorkersForDeep: 2, minPlanningWorkersForMaximum: 2, minExecutionWorkersForDeep: 3, minExecutionWorkersForMaximum: 3, minRepairWorkersForDeep: 2, minRepairWorkersForMaximum: 2, minReviewWorkersForDeep: 3, minReviewWorkersForMaximum: 3, minValidationWorkersForDeep: 3, minValidationWorkersForMaximum: 3 }, statusWidgetVisible: true },
       missions: { defaultAutonomy: "approval_gated", requireValidationPerMilestone: true, autoRunAfterApproval: true, continueAcrossMilestones: true, pauseBetweenMilestones: false, clarificationMode: "always_for_nontrivial", maxClarificationQuestions: 5, planningDepth: "deep", useSubagentsBeforeClarification: true, autoRepairReviewFailures: true, reviewRetryMode: "safe_only", maxReviewRetriesPerMission: 3, autoRepairValidationFailures: true, validationRetryMode: "safe_only", maxValidationRetriesPerMilestone: 3, maxValidationRetriesPerMission: 8, finalValidationEnabled: true, autoRepairFinalValidationFailures: true, maxFinalValidationRetries: 2, subagentPolicy: "forced", minWorkersForDeep: 3, minWorkersForMaximum: 3 },
       subagents: { planningPolicy: "forced", executionPolicy: "forced", repairPolicy: "forced", reviewPolicy: "forced", validationPolicy: "forced", autoUseDuringPlanning: true, autoUseDuringExecution: true, autoUseDuringRepair: true, autoUseDuringReview: true, autoUseDuringValidation: true, minPlanningWorkersForDeep: 3, minPlanningWorkersForMaximum: 3, minExecutionWorkersForDeep: 3, minExecutionWorkersForMaximum: 3, minRepairWorkersForDeep: 2, minRepairWorkersForMaximum: 2, minReviewWorkersForDeep: 3, minReviewWorkersForMaximum: 3, minValidationWorkersForDeep: 3, minValidationWorkersForMaximum: 3, allowBackgroundSubagents: true },
     },
@@ -968,7 +994,7 @@ export function builtInWorkflowPresets(): Record<string, WorkflowPresetBundle> {
       description: "Highest-rigor end-to-end workflow with strongest clarification, automatic review/validation, final mission validation, aggressive in-scope repair, and maximum worker teams.",
       planning: { depth: "maximum", clarificationMode: "always_for_nontrivial", maxClarificationQuestions: 5, interactiveClarificationEnabled: true, clarificationQualityGate: true, useSubagentsBeforeClarification: true },
       workflow: { offerReviewerBeforeExecute: false, autoRunReviewerBeforeExecute: true, offerValidationAfterExecute: true, autoRunValidationAfterExecute: true, validateAfterExecution: true, requirePlanApprovalBeforeExecute: false, requireApprovalBeforeExecution: false, autoRepairReviewFailures: true, autoRepairValidationFailures: true, reviewRetryMode: "aggressive_within_scope", validationRetryMode: "aggressive_within_scope", maxReviewRetriesPerPlan: 5, maxReviewRetriesPerWorkflow: 8, maxValidationRetriesPerPlan: 5, maxValidationRetriesPerWorkflow: 8, pauseAfterReviewFailure: false, pauseAfterValidationFailure: false, planProgressEnabled: true, planRuntimeEnabled: true, planShowProgressBar: true },
-      standard: { enabled: true, autoTodoEnabled: true, todoProgressVisible: true, todoTriggerMode: "auto", clarificationEnabled: true, clarificationMode: "auto", maxClarificationQuestions: 2, interactiveClarificationEnabled: true, clarificationTiming: "after_initial_analysis", clarificationQualityGate: true, allowClarificationWithoutAnalysis: false, useSubagentsBeforeClarification: true, allowSubagents: true, subagentScope: "user", subagents: { planningPolicy: "forced", executionPolicy: "forced", repairPolicy: "forced", reviewPolicy: "forced", validationPolicy: "forced", autoUseDuringPlanning: true, autoUseDuringExecution: true, autoUseDuringRepair: true, autoUseDuringReview: true, autoUseDuringValidation: true, minPlanningWorkersForDeep: 3, minPlanningWorkersForMaximum: 3, minExecutionWorkersForDeep: 4, minExecutionWorkersForMaximum: 4, minRepairWorkersForDeep: 3, minRepairWorkersForMaximum: 3, minReviewWorkersForDeep: 4, minReviewWorkersForMaximum: 4, minValidationWorkersForDeep: 4, minValidationWorkersForMaximum: 4 }, statusWidgetVisible: true, useSharedExecutorModel: true, useStandardSpecificModels: false, modelRole: "executor" },
+      standard: { enabled: true, autoTodoEnabled: true, todoProgressVisible: true, todoTriggerMode: "auto", clarificationEnabled: true, clarificationMode: "auto", maxClarificationQuestions: 2, interactiveClarificationEnabled: true, clarificationTiming: "after_initial_analysis", clarificationQualityGate: true, allowClarificationWithoutAnalysis: false, useSubagentsBeforeClarification: true, allowSubagents: true, subagentScope: "user", subagents: { planningPolicy: "forced", executionPolicy: "forced", repairPolicy: "forced", reviewPolicy: "forced", validationPolicy: "forced", autoUseDuringPlanning: true, autoUseDuringExecution: true, autoUseDuringRepair: true, autoUseDuringReview: true, autoUseDuringValidation: true, minPlanningWorkersForDeep: 3, minPlanningWorkersForMaximum: 3, minExecutionWorkersForDeep: 4, minExecutionWorkersForMaximum: 4, minRepairWorkersForDeep: 3, minRepairWorkersForMaximum: 3, minReviewWorkersForDeep: 4, minReviewWorkersForMaximum: 4, minValidationWorkersForDeep: 4, minValidationWorkersForMaximum: 4 }, statusWidgetVisible: true },
       missions: { defaultAutonomy: "supervised_auto", requireValidationPerMilestone: true, autoRunAfterApproval: true, continueAcrossMilestones: true, pauseBetweenMilestones: false, clarificationMode: "always_for_nontrivial", maxClarificationQuestions: 6, planningDepth: "maximum", useSubagentsBeforeClarification: true, autoRepairReviewFailures: true, reviewRetryMode: "aggressive_within_scope", maxReviewRetriesPerMission: 5, autoRepairValidationFailures: true, validationRetryMode: "aggressive_within_scope", maxValidationRetriesPerMilestone: 4, maxValidationRetriesPerMission: 12, finalValidationEnabled: true, autoRepairFinalValidationFailures: true, maxFinalValidationRetries: 4, subagentPolicy: "forced", minWorkersForDeep: 4, minWorkersForMaximum: 4 },
       subagents: { planningPolicy: "forced", executionPolicy: "forced", repairPolicy: "forced", reviewPolicy: "forced", validationPolicy: "forced", autoUseDuringPlanning: true, autoUseDuringExecution: true, autoUseDuringRepair: true, autoUseDuringReview: true, autoUseDuringValidation: true, minPlanningWorkersForDeep: 4, minPlanningWorkersForMaximum: 4, minExecutionWorkersForDeep: 4, minExecutionWorkersForMaximum: 4, minRepairWorkersForDeep: 3, minRepairWorkersForMaximum: 3, minReviewWorkersForDeep: 4, minReviewWorkersForMaximum: 4, minValidationWorkersForDeep: 4, minValidationWorkersForMaximum: 4, allowBackgroundSubagents: true },
     },
@@ -977,35 +1003,43 @@ export function builtInWorkflowPresets(): Record<string, WorkflowPresetBundle> {
 
 function normalizeWorkflowPresetBundle(preset: WorkflowPresetBundle): WorkflowPresetBundle & { standard: WorkflowSettings["standard"] } {
   const defaults = defaultWorkflowSettings();
-  const { models: _ignoredPresetStandardModels, ...standardWithoutModels } = (preset.standard ?? {}) as Partial<WorkflowSettings["standard"]>;
+  const standardWithoutModelRouting = standardSettingsWithoutModelRouting(preset.standard as Partial<WorkflowSettings["standard"]> | undefined);
   return {
     ...preset,
     // Legacy custom presets saved before Standard Mode do not contain a
     // standard section. Hydrate it from safe defaults during apply so stale
     // Standard settings from the previously active preset cannot leak forward.
-    standard: normalizeStandardSettings(defaults, defaults.standard, standardWithoutModels as Partial<WorkflowSettings>["standard"] | undefined),
+    standard: normalizeStandardSettings(defaults, defaults.standard, standardWithoutModelRouting as Partial<WorkflowSettings>["standard"] | undefined),
+    missions: missionSettingsWithoutModelRouting(preset.missions as Partial<WorkflowSettings["missions"]> | undefined),
   };
 }
 
 function workflowPresetBundleWithoutUserOwnedSettings(preset: WorkflowPresetBundle): WorkflowPresetBundle {
   const { context: _ignoredContext, ...presetWithoutContext } = preset as WorkflowPresetBundle & { context?: unknown };
-  const { models: _ignoredStandardModels, ...standardWithoutModels } = (preset.standard ?? {}) as Partial<WorkflowSettings["standard"]>;
-  const { models: _ignoredMissionModels, ...missionsWithoutModels } = (preset.missions ?? {}) as Partial<WorkflowSettings["missions"]>;
   return {
     ...presetWithoutContext,
-    standard: { ...standardWithoutModels },
-    missions: { ...missionsWithoutModels },
+    standard: { ...standardSettingsWithoutModelRouting(preset.standard as Partial<WorkflowSettings["standard"]> | undefined) },
+    missions: { ...missionSettingsWithoutModelRouting(preset.missions as Partial<WorkflowSettings["missions"]> | undefined) },
   };
 }
 
 function applyPresetBundle(settings: WorkflowSettings, preset: WorkflowPresetBundle, name: string): void {
   const defaults = defaultWorkflowSettings();
   const normalized = normalizeWorkflowPresetBundle(preset);
-  const currentStandardModels = settings.standard.models;
+  const currentStandardModelRouting = {
+    models: settings.standard.models,
+    useSharedExecutorModel: settings.standard.useSharedExecutorModel,
+    useStandardSpecificModels: settings.standard.useStandardSpecificModels,
+    modelRole: settings.standard.modelRole,
+  };
+  const currentMissionModelRouting = {
+    models: settings.missions.models,
+    useMissionSpecificModels: settings.missions.useMissionSpecificModels,
+  };
   if (normalized.planning) settings.planning = { ...settings.planning, ...normalized.planning };
   if (normalized.workflow) settings.workflow = { ...settings.workflow, ...normalized.workflow };
-  settings.standard = { ...defaults.standard, ...normalized.standard, subagents: { ...(defaults.standard.subagents ?? {}), ...(normalized.standard?.subagents ?? {}) }, models: currentStandardModels };
-  if (normalized.missions) settings.missions = { ...settings.missions, ...normalized.missions, models: settings.missions.models };
+  settings.standard = { ...defaults.standard, ...normalized.standard, subagents: { ...(defaults.standard.subagents ?? {}), ...(normalized.standard?.subagents ?? {}) }, ...currentStandardModelRouting };
+  if (normalized.missions) settings.missions = { ...settings.missions, ...normalized.missions, ...currentMissionModelRouting };
   if (normalized.subagents) settings.subagents = { ...settings.subagents, ...normalized.subagents };
   if (normalized.ui) settings.ui = { ...settings.ui, ...normalized.ui };
   settings.presets = { ...(settings.presets ?? {}), activePreset: name, items: { ...(settings.presets?.items ?? {}) } };
@@ -1017,34 +1051,34 @@ export function normalizeWorkflowPresetName(name: string): string {
 
 export function workflowPresetCatalog(settings: WorkflowSettings = loadGlobalSettings()): Record<string, WorkflowPresetBundle> {
   const items = { ...(settings.presets?.items ?? {}) };
-  delete items[WORKFLOW_MANUAL_PRESET];
+  delete items[WORKFLOW_CUSTOM_PRESET_MARKER];
   return { ...builtInWorkflowPresets(), ...items };
 }
 
 export function workflowPresetNames(settings: WorkflowSettings = loadGlobalSettings()): string[] {
   const builtIns = ["simple", "standard", "deep", "maximum"];
   const custom = Object.keys(settings.presets?.items ?? {}).sort((a, b) => a.localeCompare(b));
-  return [...builtIns, ...custom.filter((name) => name !== WORKFLOW_MANUAL_PRESET && !builtIns.includes(name))];
+  return [...builtIns, ...custom.filter((name) => name !== WORKFLOW_CUSTOM_PRESET_MARKER && !builtIns.includes(name))];
 }
 
 export function workflowPresetLabel(name: string, preset?: WorkflowPresetBundle): string {
-  if (name === WORKFLOW_MANUAL_PRESET) return "Manual settings";
+  if (name === WORKFLOW_CUSTOM_PRESET_MARKER) return "No active preset";
   return preset?.displayName?.trim() || name;
 }
 
 export function workflowPresetTitle(name: string, preset?: WorkflowPresetBundle): string {
-  if (name === WORKFLOW_MANUAL_PRESET) return "Manual settings — no saved preset active";
+  if (name === WORKFLOW_CUSTOM_PRESET_MARKER) return "No active preset";
   return preset?.displayName?.trim() || name;
 }
 
 export function activeWorkflowPresetLabel(settings: WorkflowSettings): string {
-  const active = settings.presets?.activePreset ?? WORKFLOW_MANUAL_PRESET;
-  if (active === WORKFLOW_MANUAL_PRESET) return "Manual settings";
+  const active = settings.presets?.activePreset ?? WORKFLOW_CUSTOM_PRESET_MARKER;
+  if (active === WORKFLOW_CUSTOM_PRESET_MARKER) return "No active preset";
   return workflowPresetLabel(active, workflowPresetCatalog(settings)[active]);
 }
 
 function assertValidCustomPresetName(name: string, safe: string): void {
-  if (safe === WORKFLOW_MANUAL_PRESET) throw new Error(`Reserved workflow preset name: ${name}. Use a specific preset name instead of custom.`);
+  if (safe === WORKFLOW_CUSTOM_PRESET_MARKER) throw new Error(`Reserved workflow preset name: ${name}. Use a specific preset name instead of custom.`);
 }
 
 function sectionCoverage(section: unknown): "configured" | "missing" {
@@ -1052,9 +1086,9 @@ function sectionCoverage(section: unknown): "configured" | "missing" {
 }
 
 export function workflowPresetDiagnostics(name: string, preset?: WorkflowPresetBundle): string[] {
-  if (name === WORKFLOW_MANUAL_PRESET) return ["Diagnostic: manual settings marker; no saved preset bundle is active."];
+  if (name === WORKFLOW_CUSTOM_PRESET_MARKER) return ["Diagnostic: reserved custom marker; no saved preset bundle is active."];
   const lines: string[] = [];
-  if (normalizeWorkflowPresetName(name) === WORKFLOW_MANUAL_PRESET) lines.push("Diagnostic: reserved-name collision; rename this preset before use.");
+  if (normalizeWorkflowPresetName(name) === WORKFLOW_CUSTOM_PRESET_MARKER) lines.push("Diagnostic: reserved-name collision; rename this preset before use.");
   const standard = sectionCoverage(preset?.standard);
   const planning = sectionCoverage(preset?.planning);
   const workflow = sectionCoverage(preset?.workflow);
@@ -1077,15 +1111,15 @@ export function workflowPresetPickerLabel(name: string, preset?: WorkflowPresetB
   const summaries: Record<string, string> = {
     simple: "fast autonomous validation, low repair retries, 1-worker phases",
     standard: "balanced autonomous validation and safe repair",
-    deep: "careful autonomous review, validation, and final mission validation",
+    deep: "careful manual review, autonomous validation, and final mission validation",
     maximum: "maximum autonomous rigor with bounded aggressive repair",
   };
   return summaries[name] ? `${title} — ${summaries[name]}` : title;
 }
 
 export function workflowPresetMeaningLines(name: string, preset?: WorkflowPresetBundle): string[] {
-  if (name === WORKFLOW_MANUAL_PRESET) return [
-    "Name: Manual settings — no saved preset active",
+  if (name === WORKFLOW_CUSTOM_PRESET_MARKER) return [
+    "Name: No active preset",
     "Purpose: uses the current workflow settings exactly as saved; this is not a built-in speed or rigor profile.",
     "Applies to: the Standard, Plan, Mission, and shared sub-agent settings currently saved in workflow-settings.json.",
     "Standard Mode: uses the saved Standard Mode To Do, clarification, continuation, and sub-agent settings.",
@@ -1122,7 +1156,7 @@ export function workflowPresetMeaningLines(name: string, preset?: WorkflowPreset
       "Purpose: higher-rigor built-in profile for broad, risky, or codebase-heavy work that still should not stall unnecessarily.",
       "Applies to: Standard Mode, Plan Mode, Mission Mode, and shared sub-agent intensity.",
       "Standard Mode: requires a To Do for substantive work, uses stronger clarification, resumes automatically after answers, and uses deeper Standard worker coverage.",
-      "Plan Mode: uses deep planning, clarifies non-trivial work, avoids a second execution-approval stop after the initial approval, runs automatic review and validation, and retries safe repairs.",
+      "Plan Mode: uses deep planning, clarifies non-trivial work, avoids a second execution-approval stop after the initial approval, leaves review manual/optional, runs validation automatically, and retries safe repairs.",
       "Mission Mode: starts after approval, auto-runs after approval, continues milestones without pause, keeps milestone validation on, and enables final comprehensive validation.",
       "Shared sub-agents: forces larger teams across planning, execution, repair, review, and validation.",
       "Does not change: models/providers/API keys/auth/session/runtime state/shared compaction settings.",
@@ -1160,11 +1194,11 @@ export function renderWorkflowPresetCard(name: string, preset?: WorkflowPresetBu
 }
 
 export function renderActiveWorkflowPresetSummary(settings: WorkflowSettings): string {
-  const active = settings.presets?.activePreset ?? WORKFLOW_MANUAL_PRESET;
-  const preset = active === WORKFLOW_MANUAL_PRESET ? undefined : workflowPresetCatalog(settings)[active];
+  const active = settings.presets?.activePreset ?? WORKFLOW_CUSTOM_PRESET_MARKER;
+  const preset = active === WORKFLOW_CUSTOM_PRESET_MARKER ? undefined : workflowPresetCatalog(settings)[active];
   const lines = workflowPresetMeaningLines(active, preset);
   lines.push("Effective Settings: detailed current values are listed below.");
-  if (active !== WORKFLOW_MANUAL_PRESET) lines.push(`Reapply Command: /workflow presets apply ${active}`);
+  if (active !== WORKFLOW_CUSTOM_PRESET_MARKER) lines.push(`Reapply Command: /workflow presets apply ${active}`);
   return lines.join("\n");
 }
 
@@ -1184,10 +1218,10 @@ export function resolveWorkflowPresetName(settings: WorkflowSettings, input: str
 
 export function renderWorkflowPresets(settings: WorkflowSettings = loadGlobalSettings()): string {
   const catalog = workflowPresetCatalog(settings);
-  const active = settings.presets?.activePreset ?? WORKFLOW_MANUAL_PRESET;
+  const active = settings.presets?.activePreset ?? WORKFLOW_CUSTOM_PRESET_MARKER;
   const names = workflowPresetNames(settings);
-  const cards = [renderWorkflowPresetCard(WORKFLOW_MANUAL_PRESET, undefined, active === WORKFLOW_MANUAL_PRESET), ...names.map((name) => renderWorkflowPresetCard(name, catalog[name], name === active))];
-  return `# Workflow Presets\n\nActive Preset: ${activeWorkflowPresetLabel(settings)}\nShortcut: Ctrl+Shift+U cycles saved presets while Plan/Mission/Standard Mode is active\nSelector: /workflow presets\n\n${cards.join("\n\n") || "No presets available."}\n\nQuick commands:\n- /workflow presets list\n- /workflow presets apply <name>\n- /workflow presets next\n- /workflow presets prev\n- /workflow presets manual\n- /workflow presets save <name>\n- /workflow presets create <name> from simple|standard|deep|maximum\n- /workflow presets edit <name>\n- /workflow presets rename <old> to <new>\n- /workflow presets delete <name>\n\nBuilt-in presets are package-defined and sync with the extension. User-named custom presets are saved entries in workflow-settings.json and stay in hotkey cycling. Manual settings is only the no-saved-preset-active marker; extension updates preserve workflow-settings.json and do not overwrite custom presets. Presets adjust workflow behavior only and preserve model/provider choices, API keys, auth/session files, runtime workflow state, and shared compaction settings.`;
+  const cards = [renderWorkflowPresetCard(WORKFLOW_CUSTOM_PRESET_MARKER, undefined, active === WORKFLOW_CUSTOM_PRESET_MARKER), ...names.map((name) => renderWorkflowPresetCard(name, catalog[name], name === active))];
+  return `# Workflow Presets\n\nActive Preset: ${activeWorkflowPresetLabel(settings)}\nShortcut: Ctrl+Shift+U cycles saved presets while Plan/Mission/Standard Mode is active\nSelector: /workflow presets\n\n${cards.join("\n\n") || "No presets available."}\n\nQuick commands:\n- /workflow presets list\n- /workflow presets apply <name>\n- /workflow presets next\n- /workflow presets prev\n- /workflow presets save <name>\n- /workflow presets create <name> from simple|standard|deep|maximum\n- /workflow presets edit <name>\n- /workflow presets rename <old> to <new>\n- /workflow presets delete <name>\n\nBuilt-in presets are package-defined and sync with the extension. User-named custom presets are saved entries in workflow-settings.json and stay in hotkey cycling. The reserved custom marker only means no built-in or user-named preset is active. Extension updates preserve workflow-settings.json and do not overwrite custom presets. Presets adjust workflow behavior only and preserve model/provider choices, API keys, auth/session files, runtime workflow state, and shared compaction settings.`;
 }
 
 export function applyWorkflowPreset(cwd: string, requestedScope: WorkflowSettingsScope | undefined, name: string): SettingsWriteResult {
@@ -1205,14 +1239,13 @@ export function saveCurrentWorkflowPreset(cwd: string, requestedScope: WorkflowS
   assertValidCustomPresetName(name, safe);
   return updateSettings(cwd, requestedScope, (settings) => {
     settings.presets = { ...(settings.presets ?? {}), activePreset: safe, items: { ...(settings.presets?.items ?? {}) } };
-    const { models: _ignoredStandardModels, ...standardPreset } = settings.standard;
     settings.presets.items![safe] = {
       displayName: name.trim(),
       description: "Custom saved workflow preset.",
       planning: { ...settings.planning },
       workflow: { ...settings.workflow },
-      standard: { ...standardPreset },
-      missions: { ...settings.missions, models: undefined } as WorkflowPresetBundle["missions"],
+      standard: { ...standardSettingsWithoutModelRouting(settings.standard) } as WorkflowPresetBundle["standard"],
+      missions: { ...missionSettingsWithoutModelRouting(settings.missions) } as WorkflowPresetBundle["missions"],
       subagents: { ...settings.subagents },
       ui: { showWorkflowStatus: settings.ui.showWorkflowStatus, showPlanModeIndicator: settings.ui.showPlanModeIndicator, planTopWidgetVisible: settings.ui.planTopWidgetVisible, planBottomWidgetVisible: settings.ui.planBottomWidgetVisible, missionTopWidgetVisible: settings.ui.missionTopWidgetVisible, missionBottomWidgetVisible: settings.ui.missionBottomWidgetVisible, workflowTheme: settings.ui.workflowTheme, workflowThemeOverrides: settings.ui.workflowThemeOverrides, startupVisual: settings.ui.startupVisual, startupVisualOnSessionStart: settings.ui.startupVisualOnSessionStart, customBrandEnabled: settings.ui.customBrandEnabled, customBrandText: settings.ui.customBrandText },
     };
@@ -1256,7 +1289,7 @@ export function deleteWorkflowPreset(cwd: string, requestedScope: WorkflowSettin
     if (!resolved || builtInWorkflowPresets()[resolved]) throw new Error(`Cannot delete unknown or built-in workflow preset: ${name}`);
     const items = { ...(settings.presets?.items ?? {}) };
     delete items[resolved];
-    settings.presets = { ...(settings.presets ?? {}), activePreset: settings.presets?.activePreset === resolved ? "custom" : settings.presets?.activePreset, items };
+    settings.presets = { ...(settings.presets ?? {}), activePreset: settings.presets?.activePreset === resolved ? WORKFLOW_CUSTOM_PRESET_MARKER : settings.presets?.activePreset, items };
   }, { preserveActivePreset: true });
 }
 
@@ -1413,7 +1446,7 @@ export function workflowSettingsConsistencyDiagnostics(settings: WorkflowSetting
   if (settings.missions.autoRepairReviewFailures !== false && settings.missions.reviewRetryMode === "off") {
     diagnostics.push("mission review auto-repair is enabled but reviewRetryMode is off — override to safe_only will be applied at runtime; set reviewRetryMode explicitly to avoid confusion");
   }
-  const activePreset = settings.presets?.activePreset ?? WORKFLOW_MANUAL_PRESET;
+  const activePreset = settings.presets?.activePreset ?? WORKFLOW_CUSTOM_PRESET_MARKER;
   if ((activePreset === "standard" || activePreset === "deep" || activePreset === "maximum")
     && (settings.missions.autoRepairReviewFailures === false || settings.missions.reviewRetryMode === "off" || (settings.missions.maxReviewRetriesPerMission ?? 0) <= 0)) {
     diagnostics.push(`active preset ${activePreset} expects Mission review repair to be available, but effective Mission review repair is disabled/off/zero; reapply the preset or update missions.autoRepairReviewFailures, missions.reviewRetryMode, and missions.maxReviewRetriesPerMission`);