@opengsd/gsd-pi 1.1.1-dev.b2556262 → 1.2.0-dev.0b870afa

package/src/resources/extensions/gsd/auto-dispatch.ts

@@ -20,6 +20,8 @@ import { getUatBrowserToolSupportError, type UatType } from "./uat-policy.js";
 import {
   isDbAvailable,
   getMilestoneSlices,
+  getMilestoneSliceSummaries,
+  getClosedSliceIds,
   getPendingGatesForTurn,
   markPendingGatesOmittedForTurn,
   getMilestone,
@@ -47,7 +49,6 @@ import {
   buildTaskFileName,
   gsdProjectionRoot,
 } from "./paths.js";
-import { parseRoadmap } from "./parsers-legacy.js";
 import { validateArtifact } from "./schemas/validate.js";
 import { existsSync, mkdirSync, readFileSync, writeFileSync, unlinkSync, readdirSync } from "node:fs";
 import { logWarning, logError } from "./workflow-logger.js";
@@ -84,13 +85,15 @@ import { selectReactiveDispatchBatch } from "./uok/execution-graph.js";
 import { getMilestonePipelineVariant } from "./milestone-scope-classifier.js";
 import { EXECUTION_ENTRY_PHASES, hasFinalizedMilestoneContext } from "./uok/plan-v2.js";
 import { isAutoActive } from "./auto.js";
-import { markDepthVerified } from "./bootstrap/write-gate.js";
+// Host adapter explicitly: auto-dispatch runs in the extension host, and the
+// ambient write-gate exports env-sniff the adapter per call (they are reserved
+// for the workflow MCP child's dynamic-import surface).
+import { hostWriteGateAdapter } from "./bootstrap/write-gate.js";
 import { ensureWorkflowPreferencesCaptured } from "./planning-depth.js";
 import { MILESTONE_ID_RE } from "./milestone-ids.js";
-import {
-  getWorkflowTransportSupportError,
-  getRequiredWorkflowToolsForAutoUnit,
-} from "./workflow-mcp.js";
+import { resolveWorkflowMcpProjectRoot } from "./workflow-mcp.js";
+import { getUnitWorkflowDispatchReadinessError } from "./tool-contract.js";
+import { prepareBrowserDaemonForUat } from "./browser-daemon-auto-prep.js";
 import {
   PROJECT_RESEARCH_INFLIGHT_MARKER,
 } from "./project-research-policy.js";
@@ -112,9 +115,9 @@ import { runTurnGitAction } from "./git-service.js";
 import { parseUnitId } from "./unit-id.js";
 import { resolveExpectedArtifactPath } from "./auto-artifact-paths.js";
 import {
-  checkCloseoutConsistencyGate,
-  formatCloseoutConsistencyBlock,
-} from "./closeout-consistency-gate.js";
+  formatCloseoutProofBlock,
+  proveMilestoneCloseout,
+} from "./milestone-closeout-proof.js";
 
 // ─── Types ────────────────────────────────────────────────────────────────
 
@@ -153,6 +156,8 @@ export interface DispatchContext {
   sessionProvider?: string;
   /** Active tools in the current session, used for transport preflight checks. */
   activeTools?: string[];
+  /** Registered tools in the current session, used for run-uat tools re-scoped at dispatch. */
+  registeredTools?: string[];
   /** Session model base URL, used for transport preflight checks. */
   sessionBaseUrl?: string;
   /** Session model auth mode, used for transport preflight checks. */
@@ -488,27 +493,12 @@ function persistSliceAssessmentBackfill(
 }
 
 function backfillMissingAssessmentsFromSummaries(basePath: string, mid: string): void {
-  const completedSliceIds = new Set<string>();
-  if (isDbAvailable()) {
-    for (const slice of getMilestoneSlices(mid)) {
-      if (slice.status === "complete" || slice.status === "done") {
-        completedSliceIds.add(slice.id);
-      }
-    }
-  } else {
-    const roadmapFile = resolveMilestoneFile(basePath, mid, "ROADMAP");
-    if (!roadmapFile) return;
-    try {
-      const roadmap = parseRoadmap(readFileSync(roadmapFile, "utf-8"));
-      for (const slice of roadmap.slices) {
-        if (slice.done) completedSliceIds.add(slice.id);
-      }
-    } catch {
-      return;
-    }
-  }
-
-  for (const sliceId of completedSliceIds) {
+  // DB-authoritative (ADR-017): no markdown fallback. Without DB rows there
+  // is nothing to backfill.
+  if (!isDbAvailable()) return;
+  // Canonical closed vocabulary (complete/done/skipped/closed) — a skipped or
+  // closed slice with a SUMMARY gets the same assessment backfill treatment.
+  for (const sliceId of getClosedSliceIds(mid)) {
     const summaryPath = resolveSliceFile(basePath, mid, sliceId, "SUMMARY");
     if (!summaryPath || !existsSync(summaryPath)) continue;
 
@@ -694,7 +684,7 @@ export const DISPATCH_RULES: DispatchRule[] = [
       // deadlock. Deep planning is still user-driven even inside auto-mode,
       // so it must wait for explicit approval instead of taking this bypass.
       if (shouldBypassMilestoneDepthGateInAuto(prefs)) {
-        markDepthVerified(mid, basePath);
+        hostWriteGateAdapter.markDepthVerified(mid, basePath);
       }
       return {
         action: "dispatch",
@@ -734,7 +724,17 @@ export const DISPATCH_RULES: DispatchRule[] = [
   },
   {
     name: "run-uat (post-completion)",
-    match: async ({ state, mid, basePath, prefs, sessionProvider, sessionAuthMode, activeTools, sessionBaseUrl }) => {
+    match: async ({
+      state,
+      mid,
+      basePath,
+      prefs,
+      sessionProvider,
+      sessionAuthMode,
+      activeTools,
+      registeredTools,
+      sessionBaseUrl,
+    }) => {
       const needsRunUat = await checkNeedsRunUat(basePath, mid, state, prefs);
       if (!needsRunUat) return null;
       const { sliceId, uatType } = needsRunUat;
@@ -742,23 +742,38 @@ export const DISPATCH_RULES: DispatchRule[] = [
       // Transport preflight: verify required MCP tools are actually connected
       // before consuming a retry attempt. Fixes tool-starved sessions burning
       // all MAX_UAT_ATTEMPTS before stopping (#477).
-      const transportError = getWorkflowTransportSupportError(
-        sessionProvider,
-        getRequiredWorkflowToolsForAutoUnit("run-uat"),
-        { projectRoot: basePath, surface: "auto-mode", unitType: "run-uat", authMode: sessionAuthMode, baseUrl: sessionBaseUrl, activeTools },
-      );
+      const transportError = getUnitWorkflowDispatchReadinessError({
+        provider: sessionProvider,
+        projectRoot: basePath,
+        surface: "auto-mode",
+        unitType: "run-uat",
+        authMode: sessionAuthMode,
+        baseUrl: sessionBaseUrl,
+        activeTools,
+      });
       if (transportError) {
         return { action: "stop" as const, reason: transportError, level: "warning" as const };
       }
       const browserToolError = getUatBrowserToolSupportError({
         uatType,
         activeTools,
+        registeredTools,
         milestoneId: mid,
         sliceId,
       });
       if (browserToolError) {
         return { action: "stop" as const, reason: browserToolError, level: "warning" as const };
       }
+      const browserDaemonError = prepareBrowserDaemonForUat({
+        uatType,
+        sessionProvider,
+        sessionAuthMode,
+        sessionBaseUrl,
+        projectRoot: resolveWorkflowMcpProjectRoot(basePath),
+      });
+      if (browserDaemonError) {
+        return { action: "stop" as const, reason: browserDaemonError, level: "warning" as const };
+      }
 
       // Cap run-uat dispatch attempts to prevent infinite replay (#3624).
       // Check before incrementing so an exhausted counter cannot create a
@@ -795,23 +810,10 @@ export const DISPATCH_RULES: DispatchRule[] = [
       // Only applies when UAT dispatch is enabled
       if (!prefs?.uat_dispatch) return null;
 
-      // DB-first: prefer closed slices from DB; fall back to ROADMAP on disk.
-      let closedSliceIds: string[];
-      if (isDbAvailable()) {
-        closedSliceIds = getMilestoneSlices(mid)
-          .filter(s => isClosedStatus(s.status))
-          .map(s => s.id);
-      } else {
-        // Filesystem fallback for degraded / unmigrated projects.
-        // `slice.done` in the parsed ROADMAP is the disk-level closed signal.
-        const roadmapFile = resolveMilestoneFile(basePath, mid, "ROADMAP");
-        const roadmapContent = roadmapFile ? await loadFile(roadmapFile) : null;
-        if (!roadmapContent) return null;
-        const roadmap = parseRoadmap(roadmapContent);
-        closedSliceIds = roadmap.slices.filter(s => s.done).map(s => s.id);
-      }
-
-      for (const sliceId of closedSliceIds) {
+      // DB-authoritative (ADR-017): closed slices come from the DB only; the
+      // ROADMAP projection is never parsed for gate decisions.
+      if (!isDbAvailable()) return null;
+      for (const sliceId of getClosedSliceIds(mid)) {
         const result = await readUatGateVerdict(basePath, mid, sliceId);
         if (!result) continue;
         const { verdict, uatType } = result;
@@ -866,7 +868,7 @@ export const DISPATCH_RULES: DispatchRule[] = [
       // H6 fix (#4973): keep the non-deep auto-mode bypass, but do not
       // pre-verify deep planning's user-facing milestone approval gate.
       if (shouldBypassMilestoneDepthGateInAuto(prefs)) {
-        markDepthVerified(mid, basePath);
+        hostWriteGateAdapter.markDepthVerified(mid, basePath);
       }
       return {
         action: "dispatch",
@@ -1037,7 +1039,7 @@ export const DISPATCH_RULES: DispatchRule[] = [
       // H6 fix (#4973): keep the non-deep auto-mode bypass, but do not
       // pre-verify deep planning's user-facing milestone approval gate.
       if (shouldBypassMilestoneDepthGateInAuto(prefs)) {
-        markDepthVerified(mid, basePath);
+        hostWriteGateAdapter.markDepthVerified(mid, basePath);
       }
       return {
         action: "dispatch",
@@ -1115,13 +1117,11 @@ export const DISPATCH_RULES: DispatchRule[] = [
       // behavior.
       if (await getMilestonePipelineVariant(mid) === "trivial") return null;
 
-      // Load roadmap to find all slices
-      const roadmapFile =
-        resolveExistingExpectedArtifact("plan-milestone", mid, basePath) ??
-        resolveMilestoneFile(basePath, mid, "ROADMAP");
-      const roadmapContent = roadmapFile ? await loadFile(roadmapFile) : null;
-      if (!roadmapContent) return null;
-      const roadmap = parseRoadmap(roadmapContent);
+      // DB-authoritative slice list (ADR-017): the ROADMAP projection is
+      // never parsed for dispatch decisions. No DB / no rows → skip this rule.
+      if (!isDbAvailable()) return null;
+      const dbSlices = getMilestoneSliceSummaries(mid);
+      if (dbSlices.length === 0) return null;
 
       // Find slices that need research (no RESEARCH file, dependencies done)
       const milestoneResearchFile =
@@ -1129,14 +1129,14 @@ export const DISPATCH_RULES: DispatchRule[] = [
         resolveMilestoneFile(basePath, mid, "RESEARCH");
       const researchReadySlices: Array<{ id: string; title: string }> = [];
 
-      for (const slice of roadmap.slices) {
+      for (const slice of dbSlices) {
         if (slice.done) continue;
         // Skip S01 when milestone research exists
         if (milestoneResearchFile && slice.id === "S01") continue;
         // Skip if already has research
         if (resolveExistingExpectedArtifact("research-slice", `${mid}/${slice.id}`, basePath)) continue;
         // Skip if dependencies aren't done (check for SUMMARY files)
-        const depsComplete = (slice.depends ?? []).every((depId) =>
+        const depsComplete = slice.depends.every((depId) =>
           !!resolveExistingExpectedArtifact("complete-slice", `${mid}/${depId}`, basePath),
         );
         if (!depsComplete) continue;
@@ -1745,11 +1745,11 @@ export const DISPATCH_RULES: DispatchRule[] = [
           };
         }
         if (milestone) {
-          const closeoutGate = checkCloseoutConsistencyGate(mid, { refreshFromDisk: true });
-          if (!closeoutGate.ok) {
+          const closeoutProof = proveMilestoneCloseout(mid, { refreshFromDisk: true });
+          if (!closeoutProof.ok) {
             return {
               action: "stop",
-              reason: formatCloseoutConsistencyBlock(closeoutGate),
+              reason: formatCloseoutProofBlock(closeoutProof),
               level: "warning",
             };
           }

package/src/resources/extensions/gsd/auto-model-selection.ts

@@ -18,7 +18,7 @@ import { getSessionModelOverride } from "./session-model-override.js";
 import { logWarning } from "./workflow-logger.js";
 import { resolveUokFlags } from "./uok/flags.js";
 import { applyModelPolicyFilter } from "./uok/model-policy.js";
-import { isModelBlocked } from "./blocked-models.js";
+import { isModelBlocked, isModelTemporarilyUnavailable } from "./blocked-models.js";
 import { getRequiredWorkflowToolsForAutoUnit, isWorkflowMcpSurfaceTool } from "./workflow-mcp.js";
 
 /**
@@ -122,6 +122,14 @@ export function getToolBaselineSnapshot(pi: ExtensionAPI): string[] {
   return [...new Set([...baseline, ...live])];
 }
 
+export function getRegisteredToolSnapshot(pi: ExtensionAPI): string[] {
+  if (typeof pi.getAllTools !== "function") return getToolBaselineSnapshot(pi);
+  const names = pi.getAllTools()
+    .map((tool) => tool.name)
+    .filter((name): name is string => typeof name === "string" && name.length > 0);
+  return [...new Set(names)];
+}
+
 /**
  * Models eligible for the pre-dispatch policy gate. Prefer registry-available
  * models; when that list is empty (common after worktree resume before registry
@@ -264,6 +272,15 @@ function buildModelPolicyBlockReasons(
   }];
 }
 
+function isModelUnavailable(
+  basePath: string,
+  provider: string | undefined,
+  id: string | undefined,
+): boolean {
+  return isModelBlocked(basePath, provider, id) ||
+    isModelTemporarilyUnavailable(basePath, provider, id);
+}
+
 function restoreToolBaseline(pi: ExtensionAPI): void {
   const key = pi as unknown as object;
   const baseline = TOOL_BASELINE.get(key);
@@ -809,9 +826,9 @@ export async function selectAndApplyModel(
       // (issue #4513).  The block is persisted in .gsd/runtime/blocked-models.json
       // so it survives /gsd auto restarts — without this, the same dead model
       // gets reselected after every restart.
-      if (isModelBlocked(basePath, model.provider, model.id)) {
+      if (isModelUnavailable(basePath, model.provider, model.id)) {
         ctx.ui.notify(
-          `Skipping blocked model ${model.provider}/${model.id} (provider rejected it for this account).`,
+          `Skipping unavailable model ${model.provider}/${model.id}.`,
           "warning",
         );
         continue;
@@ -888,7 +905,7 @@ export async function selectAndApplyModel(
       for (const model of buildPolicyEligibleFallbackOrder(ctx, routingEligibleModels, autoModeStartModel)) {
         const key = `${model.provider.toLowerCase()}/${model.id.toLowerCase()}`;
         if (!policyAllowedModelKeys.has(key)) continue;
-        if (isModelBlocked(basePath, model.provider, model.id)) continue;
+        if (isModelUnavailable(basePath, model.provider, model.id)) continue;
         const ok = await pi.setModel(model, { persist: false });
         if (!ok) continue;
         appliedModel = model;
@@ -918,10 +935,10 @@ export async function selectAndApplyModel(
       autoModeStartModel,
       effectiveSessionModelOverride,
     );
-    const startBlocked = isModelBlocked(basePath, autoModeStartModel.provider, autoModeStartModel.id);
+    const startBlocked = isModelUnavailable(basePath, autoModeStartModel.provider, autoModeStartModel.id);
     if (startBlocked) {
       ctx.ui.notify(
-        `Auto-mode start model ${autoModeStartModel.provider}/${autoModeStartModel.id} is blocked for this account. Using current session model instead.`,
+        `Auto-mode start model ${autoModeStartModel.provider}/${autoModeStartModel.id} is unavailable. Using current session model instead.`,
         "warning",
       );
     } else {
@@ -932,7 +949,7 @@ export async function selectAndApplyModel(
         const ok = await pi.setModel(startModel, { persist: false });
         if (!ok) {
           const byId = availableModels.find(
-            m => m.id === autoModeStartModel.id && !isModelBlocked(basePath, m.provider, m.id),
+            m => m.id === autoModeStartModel.id && !isModelUnavailable(basePath, m.provider, m.id),
           );
           if (byId) {
             const fallbackOk = await pi.setModel(byId, { persist: false });
@@ -1033,16 +1050,36 @@ export function resolveModelId<T extends { id: string; provider: string }>(
     if (providerMatch) return providerMatch;
   }
 
-  // Prefer "anthropic" as the canonical provider for Anthropic models.
-  // Transport-specific tiebreaker (ADR-012): intentionally keys on provider,
-  // not api — we want the plain Anthropic transport when multiple are available.
-  const anthropicMatch = candidates.find(m => m.provider === "anthropic");
-  if (anthropicMatch) return anthropicMatch;
+  // Subscription/OAuth routes beat pay-per-token API when the same model ID
+  // exists on multiple providers. Order matters — first match wins.
+  for (const provider of BARE_ID_SUBSCRIPTION_PROVIDER_PRECEDENCE) {
+    const match = candidates.find(m => m.provider === provider);
+    if (match) return match;
+  }
 
   // Fall back to first non-extension candidate, or any candidate
   return candidates.find(m => !EXTENSION_PROVIDERS.has(m.provider)) ?? candidates[0];
 }
 
+/**
+ * When a bare model ID exists on multiple providers, prefer subscription/OAuth
+ * routes over pay-per-token API keys. Matches PROVIDER_ROUTES in doctor-providers
+ * but applies when *both* sides are authenticated.
+ *
+ * Order rationale:
+ * - openai-codex before github-copilot: ChatGPT-native for shared GPT IDs
+ * - google-gemini-cli before github-copilot: first-party Gemini CLI
+ * - anthropic before github-copilot: first-party Claude API/OAuth over Copilot
+ * - github-copilot before openai/google: Copilot OAuth over platform API keys
+ */
+export const BARE_ID_SUBSCRIPTION_PROVIDER_PRECEDENCE = [
+  "openai-codex",
+  "google-gemini-cli",
+  "anthropic",
+  "github-copilot",
+  "google-antigravity",
+] as const;
+
 /**
  * Flat-rate providers charge the same per request regardless of model.
  * Dynamic routing provides no cost benefit — it only degrades quality (#3453).

package/src/resources/extensions/gsd/auto-post-unit.ts

@@ -18,7 +18,7 @@ import { deriveState } from "./state.js";
 import { logWarning, logError } from "./workflow-logger.js";
 import { loadFile, parseSummary, resolveAllOverrides } from "./files.js";
 import { loadPrompt } from "./prompt-loader.js";
-import { isAwaitingUserInput } from "./user-input-boundary.js";
+import { isAwaitingUserInput } from "./consent-question.js";
 import {
   resolveMilestonePath,
   resolveSliceFile,
@@ -49,9 +49,9 @@ import { regenerateIfMissing } from "./workflow-projections.js";
 import { WorktreeStateProjection } from "./worktree-state-projection.js";
 import { createWorkspace, scopeMilestone } from "./workspace.js";
 import { normalizeWorktreePathForCompare } from "./worktree-root.js";
-import { isDbAvailable, getDbPath, refreshOpenDatabaseFromDisk, getTask, getSlice, getMilestone, getMilestoneSlices, updateTaskStatus, _getAdapter, getVerificationEvidence } from "./gsd-db.js";
-import { renderPlanCheckboxes, renderRoadmapFromDb } from "./markdown-renderer.js";
-import { parseRoadmap as parseLegacyRoadmap } from "./parsers-legacy.js";
+import { isDbAvailable, getTask, getSlice, getMilestone, getMilestoneSlices, updateTaskStatus, _getAdapter, getVerificationEvidence } from "./gsd-db.js";
+import { getWorkflowDatabasePath, refreshWorkflowDatabaseFromDisk } from "./db-workspace.js";
+import { renderPlanCheckboxes, renderRoadmapFromDb, roadmapRenderMarksSliceDone } from "./markdown-renderer.js";
 import { consumeSignal } from "./session-status-io.js";
 import {
   checkPostUnitHooks,
@@ -71,7 +71,7 @@ import {
 } from "./milestone-closeout.js";
 import type { AutoSession, SidecarItem } from "./auto/session.js";
 import { getEvidence, clearEvidenceFromDisk, isExecutionToolName } from "./safety/evidence-collector.js";
-import { validateFileChanges } from "./safety/file-change-validator.js";
+import { validateFileChanges, effectiveFileChangeAllowlist } from "./safety/file-change-validator.js";
 import { crossReferenceEvidence, type ClaimedEvidence } from "./safety/evidence-cross-ref.js";
 import { validateContent } from "./safety/content-validator.js";
 import { resolveSafetyHarnessConfig } from "./safety/safety-harness.js";
@@ -87,7 +87,7 @@ import { writeTurnGitTransaction } from "./uok/gitops.js";
 import { isClosedStatus } from "./status-guards.js";
 import { detectAbandonMilestone } from "./abandon-detect.js";
 import { getPendingGate } from "./bootstrap/write-gate.js";
-import { isDeterministicPolicyError } from "./auto-tool-tracking.js";
+import { isDeterministicPolicyError, isToolUnavailableError } from "./auto-tool-tracking.js";
 import { formatConnectedStepStack, formatPostUnitStatusCard } from "./auto-status-message.js";
 import {
   clearProjectResearchInflightMarker,
@@ -95,6 +95,7 @@ import {
 } from "./project-research-policy.js";
 import { validateArtifact } from "./schemas/validate.js";
 import { verificationRetryKey } from "./auto/verification-retry-policy.js";
+import { saveCustomVerifyRetryCounts } from "./auto/custom-verify-retry-store.js";
 import { getLedger } from "./metrics.js";
 import { getUnitCostSpikeAction, resolveUnitCostSpikeMultiplier } from "./auto-budget.js";
 import { resolveCanonicalMilestoneRoot } from "./worktree-manager.js";
@@ -618,8 +619,10 @@ export function _hasExecutionToolCallsInSessionForTest(entries: readonly unknown
       return true;
     }
 
-    if (e?.type !== "message") continue;
-    const msg = e?.message;
+    // Accept both session-manager entries ({type: "message", message}) and
+    // bare agent-end messages ({role, content}) — the auto loop passes the
+    // latter via opts.agentEndMessages.
+    const msg = e?.type === "message" ? e?.message : e;
     if (!msg || msg.role !== "assistant" || !Array.isArray(msg.content)) continue;
     for (const block of msg.content) {
       if (block?.type !== "toolCall") continue;
@@ -898,7 +901,7 @@ export const USER_DRIVEN_DEEP_UNITS = new Set([
   "discuss-milestone",
   "research-decision",
 ]);
-export { isAwaitingUserInput } from "./user-input-boundary.js";
+export { isAwaitingUserInput } from "./consent-question.js";
 
 function artifactValidationKind(unitType: string): "project" | "requirements" | null {
   if (unitType === "discuss-project") return "project";
@@ -987,11 +990,13 @@ async function repairCompleteSliceRoadmapProjection(
     return false;
   }
 
+  // Stale-render detection (ADR-017): the DB already says the slice is closed;
+  // this only checks whether the rendered ROADMAP projection reflects it, to
+  // decide whether a repair re-render is needed.
   const roadmapPath = resolveMilestoneFile(artifactBase, mid, "ROADMAP");
   if (roadmapPath && existsSync(roadmapPath)) {
     try {
-      const roadmap = parseLegacyRoadmap(readFileSync(roadmapPath, "utf-8"));
-      if (roadmap.slices.find((roadmapSlice) => roadmapSlice.id === sid)?.done) {
+      if (roadmapRenderMarksSliceDone(readFileSync(roadmapPath, "utf-8"), sid)) {
         return false;
       }
     } catch (err) {
@@ -1273,9 +1278,9 @@ export async function postUnitPreVerification(pctx: PostUnitContext, opts?: PreV
     await new Promise(r => setTimeout(r, 100));
   }
 
-  const dbPath = getDbPath();
+  const dbPath = getWorkflowDatabasePath();
   if (isDbAvailable() && dbPath && dbPath !== ":memory:") {
-    const refreshed = refreshOpenDatabaseFromDisk();
+    const refreshed = refreshWorkflowDatabaseFromDisk();
     if (!refreshed) {
       logWarning("db", "post-unit database refresh failed; derived state may be stale");
     }
@@ -1534,6 +1539,11 @@ export async function postUnitPreVerification(pctx: PostUnitContext, opts?: PreV
       if (safetyConfig.enabled) {
         const { milestone: sMid, slice: sSid, task: sTid } = parseUnitId(s.currentUnit.id);
 
+        const fileChangeAllowlist = effectiveFileChangeAllowlist(
+          safetyConfig.file_change_allowlist,
+          (prefs?.git as { manage_gitignore?: boolean } | undefined)?.manage_gitignore,
+        );
+
         // File change validation (execute-task only, after unit execution)
         if (safetyConfig.file_change_validation && s.currentUnit.type === "execute-task" && sMid && sSid && sTid) {
           try {
@@ -1553,7 +1563,7 @@ export async function postUnitPreVerification(pctx: PostUnitContext, opts?: PreV
                   files: taskRow.files,
                 })),
               );
-              const audit = validateFileChanges(s.basePath, expectedOutput, plannedFiles, safetyConfig.file_change_allowlist);
+              const audit = validateFileChanges(s.basePath, expectedOutput, plannedFiles, fileChangeAllowlist);
               if (audit && audit.violations.length > 0) {
                 const warnings = audit.violations.filter(v => v.severity === "warning");
                 for (const v of warnings) {
@@ -1575,7 +1585,7 @@ export async function postUnitPreVerification(pctx: PostUnitContext, opts?: PreV
                   s.basePath,
                   expectedOutput,
                   plannedFiles,
-                  safetyConfig.file_change_allowlist,
+                  fileChangeAllowlist,
                 );
                 if (audit && audit.violations.length > 0) {
                   const warnings = audit.violations.filter(v => v.severity === "warning");
@@ -2014,7 +2024,32 @@ export async function postUnitPreVerification(pctx: PostUnitContext, opts?: PreV
           "error",
         );
       } else if (!triggerArtifactVerified) {
-        if (s.lastToolInvocationError) {
+        if (s.lastToolInvocationError && isToolUnavailableError(s.lastToolInvocationError)) {
+          // Tool-unavailable is transient: the workflow MCP server registers
+          // its surface asynchronously, so a Unit's first call can race the
+          // registration. Retry with escalating delay, bounded at 3 attempts.
+          // ponytail: MAX constant so the guard, log, and display all agree
+          const MAX_TOOL_UNAVAIL_RETRIES = 3;
+          if (s.toolUnavailableRetries >= MAX_TOOL_UNAVAIL_RETRIES) {
+            debugLog("postUnit", { phase: "tool-unavailable-exhausted", unitType: s.currentUnit.type, unitId: s.currentUnit.id, retries: s.toolUnavailableRetries });
+            ctx.ui.notify(
+              `Tool unavailable for ${s.currentUnit.type} after ${MAX_TOOL_UNAVAIL_RETRIES} retries: ${s.lastToolInvocationError}. MCP server may not be starting — pausing auto-mode.`,
+              "error",
+            );
+            s.lastToolInvocationError = null;
+            await pauseAuto(ctx, pi);
+            return "dispatched";
+          }
+          s.toolUnavailableRetries++;
+          const delayMs = s.toolUnavailableRetries * 1000;
+          debugLog("postUnit", { phase: "tool-unavailable-retry", unitType: s.currentUnit.type, unitId: s.currentUnit.id, error: s.lastToolInvocationError, attempt: s.toolUnavailableRetries, delayMs });
+          ctx.ui.notify(
+            `Tool unavailable for ${s.currentUnit.type}: ${s.lastToolInvocationError}. Waiting ${delayMs}ms for MCP server — retry ${s.toolUnavailableRetries}/${MAX_TOOL_UNAVAIL_RETRIES}.`,
+            "warning",
+          );
+          s.lastToolInvocationError = null;
+          await new Promise(r => setTimeout(r, delayMs));
+        } else if (s.lastToolInvocationError) {
           const isUserSkip = /queued user message/i.test(s.lastToolInvocationError);
           const errMsg = isUserSkip
             ? `Tool skipped for ${s.currentUnit.type}: ${s.lastToolInvocationError}. Queued user message interrupted the turn — pausing auto-mode.`
@@ -2140,6 +2175,7 @@ export async function postUnitPreVerification(pctx: PostUnitContext, opts?: PreV
               }
             }
             s.exhaustedVerificationUnits.add(retryKey);
+            saveCustomVerifyRetryCounts(s, { logFailure: err => debugLog("postUnit", { phase: "save-verify-retries-failed", error: err instanceof Error ? err.message : String(err) }) });
             debugLog("postUnit", { phase: "artifact-verify-exhausted", unitType: s.currentUnit.type, unitId: s.currentUnit.id, attempt });
             ctx.ui.notify(
               `${failureDetails} Pausing auto-mode after ${MAX_ARTIFACT_VERIFICATION_RETRIES} retries.`,
@@ -2149,6 +2185,7 @@ export async function postUnitPreVerification(pctx: PostUnitContext, opts?: PreV
             return "dispatched";
           }
           s.verificationRetryCount.set(retryKey, attempt);
+          saveCustomVerifyRetryCounts(s, { logFailure: err => debugLog("postUnit", { phase: "save-verify-retries-failed", error: err instanceof Error ? err.message : String(err) }) });
           s.pendingVerificationRetry = {
             unitId: s.currentUnit.id,
             failureContext: `${failureDetails} (attempt ${attempt}/${MAX_ARTIFACT_VERIFICATION_RETRIES}).`,
@@ -2170,8 +2207,11 @@ export async function postUnitPreVerification(pctx: PostUnitContext, opts?: PreV
         if (s.pendingVerificationRetry?.unitId === s.currentUnit.id) {
           s.pendingVerificationRetry = null;
         }
+        s.toolUnavailableRetries = 0;
         s.verificationRetryCount.delete(retryKey);
         s.verificationRetryFailureHashes.delete(retryKey);
+        s.exhaustedVerificationUnits.delete(retryKey);
+        saveCustomVerifyRetryCounts(s, { logFailure: err => debugLog("postUnit", { phase: "save-verify-retries-failed", error: err instanceof Error ? err.message : String(err) }) });
 
         if (s.currentUnit.type === "complete-milestone") {
           const { milestone: mid } = parseUnitId(s.currentUnit.id);