@opengsd/gsd-core 1.15.0 → 1.16.0

package/scripts/gen-scripts-cli-exit.cjs

@@ -65,15 +65,40 @@ const BANNER = [
   '',
 ].join('\n');
 
-/** Compile the whole project to a throwaway outDir so the work tree is untouched. */
+/**
+ * Compile ONLY src/cli-exit.cts (and whatever it imports — tsc follows
+ * `./exit-code-registry.cjs` on its own) to a throwaway outDir, so the work
+ * tree is untouched. A whole-project `-p tsconfig.build.json` type-checks and
+ * emits every .cts file in src/ under a fixed COMPILE_TIMEOUT_MS ceiling; on a
+ * CPU-contended shared bench that whole-project compile can exceed the
+ * ceiling and get killed even though the one file this script actually needs
+ * compiles in a fraction of the time. A throwaway tsconfig that `extends`
+ * tsconfig.build.json but overrides `files`/`include` to name only
+ * cli-exit.cts keeps every compiler option (target, strict, rootDir, etc.)
+ * while scoping the type-check + emit to that file's own import closure.
+ */
 function compileToTemp() {
   const tmp = fs.mkdtempSync(path.join(os.tmpdir(), 'gsd-scripts-cli-exit-'));
+  const scopedTsconfigPath = path.join(tmp, 'tsconfig.scoped.json');
   try {
+    const scopedTsconfig = {
+      extends: path.join(REPO_ROOT, 'tsconfig.build.json'),
+      // typeRoots must be pinned to the real repo's node_modules/@types: tsc
+      // resolves a bare "types": ["node"] by walking up from the config
+      // file's own directory, and our scoped tsconfig lives outside the repo
+      // tree (in the throwaway tmp dir), so without this it cannot find
+      // @types/node at all.
+      compilerOptions: { typeRoots: [path.join(REPO_ROOT, 'node_modules', '@types')] },
+      files: [path.join(REPO_ROOT, 'src', 'cli-exit.cts')],
+      include: [],
+    };
+    fs.writeFileSync(scopedTsconfigPath, JSON.stringify(scopedTsconfig, null, 2), 'utf8');
+
     execFileSync(
       process.execPath,
       [
         path.join(REPO_ROOT, 'node_modules', 'typescript', 'bin', 'tsc'),
-        '-p', path.join(REPO_ROOT, 'tsconfig.build.json'),
+        '-p', scopedTsconfigPath,
         '--outDir', tmp,
         // A throwaway outDir must not reuse the in-tree incremental state, or
         // tsc skips emit for files it believes are already current.
@@ -87,6 +112,8 @@ function compileToTemp() {
     fs.rmSync(tmp, { recursive: true, force: true });
     const detail = [err.stdout, err.stderr].filter(Boolean).join('\n').trim();
     return { ok: false, detail };
+  } finally {
+    fs.rmSync(scopedTsconfigPath, { force: true });
   }
 }
 
@@ -182,4 +209,4 @@ function main() {
 
 if (require.main === module) process.exitCode = main();
 
-module.exports = { REASON, buildExpectedContent, OUTPUT_PATH, BANNER };
+module.exports = { REASON, buildExpectedContent, compileToTemp, OUTPUT_PATH, BANNER };

package/scripts/gen-test-timings.cjs

@@ -19,6 +19,9 @@
 //   node scripts/gen-test-timings.cjs <events.jsonl> [<events.jsonl> ...]
 //   node scripts/gen-test-timings.cjs ~/.local/state/gsd-test/runs/*/test-events-*.jsonl
 //   node scripts/gen-test-timings.cjs events.jsonl --out tests/test-timings.json
+//   node scripts/gen-test-timings.cjs --platform win32 test-timings-windows-shard*.jsonl
+//     (#5071: the per-shard exports CI uploads from its Windows conformance
+//     jobs -> tests/test-timings.win32.json, preferred by run-tests.cjs on win32)
 //
 // When several streams are supplied (multiple lanes, e.g. node22 + node24), a
 // file's recorded time is the MAX across them, not the mean: the packer exists
@@ -28,9 +31,9 @@
 // The table is ADVISORY and deliberately un-gated — there is no `--check` mode
 // and no CI lint that fails on staleness, because timing data legitimately
 // varies run to run. A file missing from the table falls back to the table's
-// median weight, so a stale table degrades chunk BALANCE gracefully instead of
-// failing the build. Regenerate it when the suite's cost profile has visibly
-// drifted, not on a schedule.
+// mean weight (see makeFileWeigher in run-tests.cjs), so a stale table
+// degrades chunk BALANCE gracefully instead of failing the build. Regenerate
+// it when the suite's cost profile has visibly drifted, not on a schedule.
 'use strict';
 
 const fs = require('fs');
@@ -40,12 +43,32 @@ const { ExitError, runMain } = require('./lib/cli-exit.cjs');
 const DEFAULT_OUT = join(__dirname, '..', 'tests', 'test-timings.json');
 const SCHEMA_VERSION = 1;
 
+// #5071: `--platform <p>` writes the table measured ON that platform, which
+// scripts/run-tests.cjs prefers there (see its makeFileWeigher). The default
+// path must match run-tests.cjs's platformTimingsPath() —
+// tests/run-tests-harness.test.cjs pins the two together. A closed set, in
+// exactly Node's own `process.platform` spelling, because the value becomes
+// part of a filename the runner looks up by that spelling.
+const PLATFORMS = ['win32', 'darwin', 'linux'];
+function platformOut(platform) {
+  return join(__dirname, '..', 'tests', `test-timings.${platform}.json`);
+}
+
 function parseArgs(argv) {
   const inputs = [];
-  let out = DEFAULT_OUT;
+  let out = null;
+  let platform = null;
   for (let i = 0; i < argv.length; i++) {
     const arg = argv[i];
-    if (arg === '--out') {
+    if (arg === '--platform' || arg.startsWith('--platform=')) {
+      if (platform !== null) return { error: 'duplicate --platform flag' };
+      const value = arg === '--platform' ? argv[++i] : arg.slice('--platform='.length);
+      if (!value || value.startsWith('-')) return { error: `--platform requires one of: ${PLATFORMS.join(', ')}` };
+      if (!PLATFORMS.includes(value)) {
+        return { error: `unknown --platform "${value}" (expected one of: ${PLATFORMS.join(', ')})` };
+      }
+      platform = value;
+    } else if (arg === '--out') {
       const value = argv[++i];
       if (!value) return { error: '--out requires a path' };
       out = value;
@@ -60,9 +83,10 @@ function parseArgs(argv) {
     }
   }
   if (inputs.length === 0) {
-    return { error: 'usage: gen-test-timings.cjs <reporter-events.jsonl> [...] [--out <path>]' };
+    return { error: 'usage: gen-test-timings.cjs <reporter-events.jsonl> [...] [--platform <win32|darwin|linux>] [--out <path>]' };
   }
-  return { inputs, out };
+  if (out === null) out = platform ? platformOut(platform) : DEFAULT_OUT;
+  return { inputs, out, platform };
 }
 
 // Fold one reporter event stream into `acc`, keeping the MAX duration seen for
@@ -180,6 +204,7 @@ function main() {
   const payload = {
     schema_version: SCHEMA_VERSION,
     generated_by: 'scripts/gen-test-timings.cjs',
+    ...(parsed.platform ? { platform: parsed.platform } : {}),
     unit: 'ms',
     sources: sources.sort(),
     file_count: acc.size,

package/scripts/lib/cli-exit.cjs

@@ -183,6 +183,17 @@ function setPendingOutcome(v) {
 function getPendingOutcome() {
     return globalThis[PENDING_OUTCOME_KEY];
 }
+/**
+ * Declare the outcome a void-returning `main()` ends with (#5170, ADR-5057 §4): writes the
+ * pending-outcome cell that `runMain` projects. Validates `name` is an outcome
+ * `projectOutcome` accepts (PASS, FAIL or a registered name) and THROWS otherwise, so a typo'd
+ * name fails at the declaration rather than at process exit. Call it AFTER `output()`, which
+ * rewrites the cell on every call. `output()` and this are the only writers of the cell.
+ */
+function declareOutcome(name) {
+    projectOutcome(name, 'v1');
+    setPendingOutcome(name);
+}
 const EXIT_CONTRACT_FLAG_PREFIX = '--exit-contract=';
 /** Scan argv for the FIRST `--exit-contract=<value>` token; undefined if absent. */
 function findExitContractFlag(argv) {
@@ -555,4 +566,5 @@ module.exports = {
     terminateNow,
     setPendingOutcome,
     getPendingOutcome,
+    declareOutcome,
 };

package/scripts/lib/macos-conformance-tier.generated.cjs

@@ -6,6 +6,7 @@
 
 module.exports = {
   MACOS_CONFORMANCE_TIER_FILES: [
+  "tests/active-workstream-store.unit.test.cjs",
   "tests/adr-612-bracket-phase-counting.test.cjs",
   "tests/adr-index-gate.test.cjs",
   "tests/adr-parser.property.test.cjs",
@@ -32,6 +33,7 @@ module.exports = {
   "tests/changeset-parse.test.cjs",
   "tests/check-contract-drift.test.cjs",
   "tests/check-predicate.test.cjs",
+  "tests/check-router-cutover-equivalence.test.cjs",
   "tests/check-ui-plan-gate.test.cjs",
   "tests/check-ui-safety-gate.test.cjs",
   "tests/check-update-config-dir.test.cjs",
@@ -44,7 +46,7 @@ module.exports = {
   "tests/codex-config-hooks.test.cjs",
   "tests/codex-config-install.test.cjs",
   "tests/codex-config.test.cjs",
-  "tests/commands.test.cjs",
+  "tests/commands.platform.test.cjs",
   "tests/commit-docs-bypass.test.cjs",
   "tests/commit-files-deletion.test.cjs",
   "tests/commit-files-pathspec.test.cjs",
@@ -61,6 +63,7 @@ module.exports = {
   "tests/default-flip-documentation-lint.test.cjs",
   "tests/discuss-phase-power.test.cjs",
   "tests/docs-parity-live-registry.test.cjs",
+  "tests/drift-whole-change-set.test.cjs",
   "tests/effort-surface-axis.test.cjs",
   "tests/effort-sync-installed-runtime.test.cjs",
   "tests/emitted-attribution.test.cjs",
@@ -75,12 +78,18 @@ module.exports = {
   "tests/features-index-gate.test.cjs",
   "tests/frontmatter.unit.test.cjs",
   "tests/gap-checker.property.test.cjs",
+  "tests/gate-api-coverage-verify-pre-escape.test.cjs",
+  "tests/gate-codebase-drift.unit.test.cjs",
+  "tests/gate-tdd-red-evidence-containment.test.cjs",
+  "tests/gate-ui-plan.unit.test.cjs",
+  "tests/gate-verify-command-paths.unit.test.cjs",
   "tests/gemini-runtime-removed.test.cjs",
   "tests/gen-context-index.test.cjs",
   "tests/gen-section-manifest.test.cjs",
   "tests/git-base-branch.test.cjs",
   "tests/graphify-query.test.cjs",
   "tests/graphify-visualization.test.cjs",
+  "tests/gsd-agent-isolation-guard.test.cjs",
   "tests/gsd-secret-read-guard.test.cjs",
   "tests/gsd-settings-advanced.test.cjs",
   "tests/gsd-statusline.test.cjs",
@@ -96,7 +105,6 @@ module.exports = {
   "tests/hooks-opt-in.test.cjs",
   "tests/host-integration.test.cjs",
   "tests/init-manager.test.cjs",
-  "tests/init.test.cjs",
   "tests/install-minimal-hooks.test.cjs",
   "tests/install-path-detection.test.cjs",
   "tests/install-regressions.test.cjs",
@@ -115,9 +123,11 @@ module.exports = {
   "tests/issue-version-gate.test.cjs",
   "tests/kimi-upgrades.test.cjs",
   "tests/lint-docs-command-form.test.cjs",
+  "tests/lint-frontmatter-scalar-broad-grep.test.cjs",
   "tests/lint-retired-runtime-name.test.cjs",
   "tests/lint-workflow-shellcheck-fetch.test.cjs",
   "tests/list-seeds.test.cjs",
+  "tests/markdown-sectionizer-xml-tags.property.test.cjs",
   "tests/markdown-sectionizer.test.cjs",
   "tests/mcp-catalog.test.cjs",
   "tests/milestone-archive.test.cjs",
@@ -142,6 +152,8 @@ module.exports = {
   "tests/phase-id-drift-guard.test.cjs",
   "tests/phase-id.test.cjs",
   "tests/phase-locator.test.cjs",
+  "tests/phase-status.test.cjs",
+  "tests/phase.platform.test.cjs",
   "tests/phase.test.cjs",
   "tests/plan-count-single-owner.test.cjs",
   "tests/plan-document.test.cjs",
@@ -162,6 +174,7 @@ module.exports = {
   "tests/profile-output.test.cjs",
   "tests/profile-pipeline.test.cjs",
   "tests/project-root.test.cjs",
+  "tests/quick-batch-command-router.stdin.test.cjs",
   "tests/quick-batch.test.cjs",
   "tests/quick-branching.test.cjs",
   "tests/quick-research.test.cjs",
@@ -183,6 +196,7 @@ module.exports = {
   "tests/roadmap-phase-fallback.test.cjs",
   "tests/roadmap-upgrade-bracket.test.cjs",
   "tests/roadmap.test.cjs",
+  "tests/run-tests-accounting.test.cjs",
   "tests/run-tests-harness.test.cjs",
   "tests/runtime-artifact-layout.test.cjs",
   "tests/runtime-identity.test.cjs",
@@ -196,11 +210,14 @@ module.exports = {
   "tests/spawn-liveness-banner.test.cjs",
   "tests/state-document.test.cjs",
   "tests/state-transition.test.cjs",
+  "tests/state.platform.test.cjs",
   "tests/state.test.cjs",
   "tests/todos-workstream-scope.test.cjs",
   "tests/tracer-bullet.test.cjs",
+  "tests/uat-complete-session.test.cjs",
   "tests/uat-predicate.test.cjs",
   "tests/uat.test.cjs",
+  "tests/ui-auditor-screenshot-block.test.cjs",
   "tests/ui-interaction-capture.test.cjs",
   "tests/ui-safety-gate.test.cjs",
   "tests/ui-spec-inventory-provenance.test.cjs",
@@ -209,6 +226,7 @@ module.exports = {
   "tests/unusable-input.test.cjs",
   "tests/update-custom-backup.test.cjs",
   "tests/user-artifact-staging.test.cjs",
+  "tests/verification-append-audit.test.cjs",
   "tests/verification-overrides.test.cjs",
   "tests/verification-status.test.cjs",
   "tests/verify-archive-dirs-live-path.test.cjs",

package/scripts/lib/ndjson-reporter.cjs

@@ -50,8 +50,9 @@
 // repo's `engines.node` requires >=24.0.0 (package.json), where both
 // contracts have been stable since Node 20.
 //
-// Kept intentionally tiny: only the five event types run-tests.cjs needs are
-// handled — `test:enqueue`/`test:dequeue` (emitted by the RUNNER as it queues
+// Kept intentionally tiny: only the event types run-tests.cjs needs are
+// handled — per-file `test:summary` (#5071, see the loop) plus these five:
+// `test:enqueue`/`test:dequeue` (emitted by the RUNNER as it queues
 // and begins each spawned test-file child, independent of whether anything
 // inside that file ever completes) plus `test:start`/`test:pass`/`test:fail`
 // (emitted per-subtest, once the child reports it). `test:dequeue` is the
@@ -94,13 +95,19 @@ module.exports = async function ndjsonEventReporter(source) {
       event.type === 'test:pass' ||
       event.type === 'test:fail'
     ) {
-      const { file, name, nesting, testNumber } = event.data || {};
+      const { file, name, nesting, testNumber, details } = event.data || {};
+      // #4031: `kind` is node:test's `details.type` ('suite' for a suite,
+      // otherwise a test). run-tests.cjs's per-chunk accounting counts the
+      // pass/fail events that are NOT suites and compares them with the
+      // registration ledger (scripts/lib/registration-ledger-preload.cjs).
+      const kind = details && typeof details.type === 'string' ? details.type : undefined;
       const line = `${JSON.stringify({
         type: event.type,
         file,
         name,
         nesting,
         testNumber,
+        kind,
         ts: Date.now(),
       })}\n`;
       try {
@@ -109,6 +116,24 @@ module.exports = async function ndjsonEventReporter(source) {
         // Best-effort: a write failure here (e.g. the events dir vanished)
         // must never crash the test run this reporter is only observing.
       }
+    } else if (event.type === 'test:summary' && typeof (event.data || {}).file === 'string') {
+      // #5071: node:test emits one `test:summary` per test FILE (its
+      // `duration_ms` is that file's wall-clock) plus a run-level one with no
+      // `file`. Only the per-file ones are recorded: run-tests.cjs exports
+      // them, when asked, as the raw input for a per-platform timing table
+      // (scripts/gen-test-timings.cjs --platform). The run-level summary is
+      // not a file measurement and is skipped here.
+      const line = `${JSON.stringify({
+        type: event.type,
+        file: event.data.file,
+        duration_ms: event.data.duration_ms,
+        ts: Date.now(),
+      })}\n`;
+      try {
+        require('fs').appendFileSync(eventsPath, line);
+      } catch {
+        // Best-effort, as above.
+      }
     }
   }
   // Falls through to an implicit `return undefined` (nully): this reporter is

package/scripts/lib/platform-conformance-tier.generated.cjs

@@ -14,7 +14,9 @@ module.exports = {
   "tests/assumption-delta-checkpoint-e2e.test.cjs",
   "tests/assumption-delta.test.cjs",
   "tests/audit-command-cutover.test.cjs",
+  "tests/audit-workstream-layouts.test.cjs",
   "tests/augment-upgrades.test.cjs",
+  "tests/blocking-guard-budget-parity.test.cjs",
   "tests/broken-windows.test.cjs",
   "tests/capability-cli.test.cjs",
   "tests/capability-command-dispatch.test.cjs",
@@ -32,6 +34,8 @@ module.exports = {
   "tests/check-gap-analysis-plan-post-e2e.test.cjs",
   "tests/check-glossary-refs.test.cjs",
   "tests/check-predicate.test.cjs",
+  "tests/check-router-cutover-equivalence.test.cjs",
+  "tests/check-router-gate-boundaries.test.cjs",
   "tests/check-tdd-review-checkpoint-e2e.test.cjs",
   "tests/check-ui-safety-gate.test.cjs",
   "tests/check-update-config-dir.test.cjs",
@@ -53,7 +57,7 @@ module.exports = {
   "tests/codex-config.test.cjs",
   "tests/codex-inherit-smoke.test.cjs",
   "tests/command-routing-hub.test.cjs",
-  "tests/commands.test.cjs",
+  "tests/commands.platform.test.cjs",
   "tests/commit-files-deletion.test.cjs",
   "tests/commit-files-pathspec.test.cjs",
   "tests/commonjs-marker.test.cjs",
@@ -63,17 +67,21 @@ module.exports = {
   "tests/config-get-default.test.cjs",
   "tests/config-loader.test.cjs",
   "tests/config-schema.property.test.cjs",
-  "tests/config.test.cjs",
+  "tests/config.platform.test.cjs",
   "tests/configured-entrypoint-validation.test.cjs",
+  "tests/context-drift.test.cjs",
   "tests/copilot-install.test.cjs",
   "tests/copilot-upgrades.test.cjs",
   "tests/core-utils.test.cjs",
   "tests/cursor-hook-workspace-roots.test.cjs",
   "tests/cursor-hooks.test.cjs",
   "tests/cursor-subagent-isolation.test.cjs",
+  "tests/debug-session-management.test.cjs",
   "tests/dispatch-identity.test.cjs",
   "tests/dispatcher.test.cjs",
   "tests/drift-detection.test.cjs",
+  "tests/drift-whole-change-set.test.cjs",
+  "tests/effort-local-install.test.cjs",
   "tests/effort-surface-axis.test.cjs",
   "tests/effort-sync-installed-runtime.test.cjs",
   "tests/emitted-ack-trailer.test.cjs",
@@ -95,6 +103,16 @@ module.exports = {
   "tests/features-index-gate.test.cjs",
   "tests/frontmatter.test.cjs",
   "tests/gap-checker.property.test.cjs",
+  "tests/gate-api-coverage-verify-pre-escape.test.cjs",
+  "tests/gate-consumer-exit-awareness.test.cjs",
+  "tests/gate-decision-coverage-verify-stderr.test.cjs",
+  "tests/gate-exit.test.cjs",
+  "tests/gate-tdd-red-evidence-containment.test.cjs",
+  "tests/gate-tdd-review-checkpoint-hardening.test.cjs",
+  "tests/gate-tdd-review-checkpoint.unit.test.cjs",
+  "tests/gate-ui-plan.unit.test.cjs",
+  "tests/gate-ui-safety.unit.test.cjs",
+  "tests/gate-verify-command-paths.unit.test.cjs",
   "tests/gemini-runtime-removed.test.cjs",
   "tests/gen-context-index.test.cjs",
   "tests/gen-health-docs.test.cjs",
@@ -118,12 +136,12 @@ module.exports = {
   "tests/helpers-cleanup.test.cjs",
   "tests/helpers-process-isolation.test.cjs",
   "tests/hermes-skills-migration.test.cjs",
+  "tests/hook-activation-probe.test.cjs",
   "tests/hooks-commonjs-marker.test.cjs",
   "tests/hooks-crash-policy.test.cjs",
   "tests/hooks-opt-in.test.cjs",
   "tests/host-integration.test.cjs",
   "tests/init-manager.test.cjs",
-  "tests/init.test.cjs",
   "tests/install-minimal-hooks.test.cjs",
   "tests/install-nested-layout.test.cjs",
   "tests/install-path-detection.test.cjs",
@@ -145,6 +163,7 @@ module.exports = {
   "tests/kimi-agent-converter.test.cjs",
   "tests/kimi-upgrades.test.cjs",
   "tests/kimi-variant-disambiguation.test.cjs",
+  "tests/launcher-homes-derivation.test.cjs",
   "tests/lint-phase-arg-assignment.test.cjs",
   "tests/lint-workflow-shellcheck-fetch.test.cjs",
   "tests/live-config-guard.test.cjs",
@@ -170,6 +189,7 @@ module.exports = {
   "tests/no-unbounded-dirname-walk.rule.test.cjs",
   "tests/no-unbounded-spawn.test.cjs",
   "tests/no-unguarded-nonportable-exec.rule.test.cjs",
+  "tests/node-compile-cache.test.cjs",
   "tests/npm-audit-baseline.test.cjs",
   "tests/npm-integrity-gate.test.cjs",
   "tests/nsegment-phase-grammar.test.cjs",
@@ -185,7 +205,7 @@ module.exports = {
   "tests/phase-completion-single-owner.test.cjs",
   "tests/phase-estimation.test.cjs",
   "tests/phase-locator.test.cjs",
-  "tests/phase.test.cjs",
+  "tests/phase.platform.test.cjs",
   "tests/phase6-capstone-conformance.test.cjs",
   "tests/pi-config-dir-env-override.test.cjs",
   "tests/plan-count-single-owner.test.cjs",
@@ -212,6 +232,7 @@ module.exports = {
   "tests/profile-pipeline.test.cjs",
   "tests/prohibition-enforcement.test.cjs",
   "tests/project-root.test.cjs",
+  "tests/quick-batch-command-router.stdin.test.cjs",
   "tests/quick-batch.test.cjs",
   "tests/quick-branching.test.cjs",
   "tests/quick-research.test.cjs",
@@ -238,15 +259,18 @@ module.exports = {
   "tests/roadmap-parser.test.cjs",
   "tests/roadmap-upgrade-bracket.test.cjs",
   "tests/roadmap.test.cjs",
+  "tests/run-tests-accounting.test.cjs",
   "tests/run-tests-harness.test.cjs",
   "tests/run-tests-temp-root.test.cjs",
   "tests/run-with-timeout.test.cjs",
   "tests/runtime-artifact-layout-surface.test.cjs",
   "tests/runtime-artifact-layout.test.cjs",
   "tests/runtime-converters.test.cjs",
+  "tests/runtime-descriptor-owner.test.cjs",
   "tests/runtime-homes-descriptor-drive.test.cjs",
   "tests/runtime-identity.test.cjs",
   "tests/runtime-launcher-parity.test.cjs",
+  "tests/schema-drift.test.cjs",
   "tests/security.test.cjs",
   "tests/settings-jsonc.test.cjs",
   "tests/sh-hook-paths.test.cjs",
@@ -258,10 +282,12 @@ module.exports = {
   "tests/skill-manifest.test.cjs",
   "tests/slug-derivation-drift-guard.test.cjs",
   "tests/state-todos-render.test.cjs",
-  "tests/state.test.cjs",
+  "tests/state.platform.test.cjs",
   "tests/teams-status.test.cjs",
   "tests/test-gate-watch-mode.test.cjs",
   "tests/todos-workstream-scope.test.cjs",
+  "tests/uat-complete-session.test.cjs",
+  "tests/ui-auditor-screenshot-block.test.cjs",
   "tests/ui-interaction-capture.test.cjs",
   "tests/undo-commit-selection-4465.test.cjs",
   "tests/undo-commit-selection.test.cjs",
@@ -270,9 +296,12 @@ module.exports = {
   "tests/unusable-input.test.cjs",
   "tests/update-custom-backup.test.cjs",
   "tests/user-artifact-staging.test.cjs",
+  "tests/verification-append-audit.test.cjs",
   "tests/verification-status.test.cjs",
   "tests/verify-archive-dirs-live-path.test.cjs",
   "tests/verify-command-grounding.test.cjs",
+  "tests/verify-lifecycle-writes-e2e.test.cjs",
+  "tests/verify-work-auto-transition.test.cjs",
   "tests/verify.test.cjs",
   "tests/windsurf-hooks-bridge.test.cjs",
   "tests/workflow-guard.test.cjs",

package/scripts/lib/registration-ledger-preload.cjs

@@ -0,0 +1,155 @@
+'use strict';
+
+// Registration ledger preload for scripts/run-tests.cjs (#4031).
+//
+// Why this exists. The runner passes `--test-force-exit` to every chunk (the
+// Windows post-test hang backstop, #1051/#869). Under process isolation each
+// test file's child streams its results to the `node --test` parent over a
+// pipe, and a force-exit can end the parent while bytes of that pipe are still
+// unread (nodejs/node#64833). The reporters never see the lost tail, so the
+// run printed a smaller pass count and exited 0: a test that executed and
+// passed (or failed) was indistinguishable from one that never existed.
+//
+// The parent cannot know what it did not receive, so the count of results the
+// child produced has to come from the child, over a channel that does not
+// share the pipe: this module is loaded into each test-file child with
+// `--require` (`--test` forwards it to the file children) and counts the leaf
+// `test:pass` / `test:fail` events the child's own reporter hands to the V8
+// serializer that frames them onto that pipe, keyed by the `file` each event
+// carries. When the child exits it appends one line per file to the file named
+// by GSD_RUN_TESTS_LEDGER_FILE. The runner compares that per-file count with
+// the leaf results its ndjson reporter actually received
+// (analyzeChunkAccounting in run-tests.cjs) and fails the chunk when
+// registered > reported.
+//
+// Why it taps the serializer and does NOT wrap `test()` / `it()`. node:test
+// stamps every test with the location of the code that called it
+// (`getCallerLocation()` in lib/internal/test_runner/harness.js) and reports
+// that as the event's `file`. A wrapper around `test` is a JS frame between the
+// test file and node:test, so every event would report THIS file as its
+// location: spec-reporter failure locations would point here and no result
+// could be matched back to its test file. Counting at the serializer adds no
+// frame to the call and counts exactly what is put on the pipe, so the count
+// and the parent's report are compared like for like: a skipped or todo test,
+// a skipped suite and a test name/only filter need no special case
+// (an excluded test emits no event on either side).
+//
+// Blind spots, stated plainly: a child that emits nothing counts 0 (no loss is
+// visible); a SIGKILLed child never runs the `exit` handler below, so it leaves no
+// ledger line; a file the parent reported but the ledger never saw is not flagged
+// per file; and the whole count depends on node:test's child reporter framing its
+// events through v8.DefaultSerializer.prototype.writeValue (true on Node 24). If a
+// Node change moves that, every count becomes 0: the runner then fails a chunk whose
+// ledger counted 0 while its reporter received results (analyzeChunkAccounting,
+// `ledgerCountedNothing`), and tests/run-tests-accounting.test.cjs pins the hook.
+//
+// Inert unless it is inside a test-file child (NODE_TEST_CONTEXT is set by
+// `node --test` for those, not for the runner parent), so requiring it anywhere
+// else is a no-op. There it does two things: it makes the child's stdout pipe
+// blocking (blockChildStdout: this is what stops `--test-force-exit` from
+// discarding queued results, the loss this accounting exists to detect), and,
+// when a ledger path was supplied, counts results. Never throws into the code it
+// observes.
+
+const fs = require('fs');
+const path = require('path');
+const { DefaultSerializer } = require('v8');
+
+const RESULT_TYPES = new Set(['test:pass', 'test:fail']);
+
+/**
+ * The file a reporter event counts toward, or null when the event is not a leaf test result the
+ * runner's ndjson reporter would record (anything but a pass/fail, a suite, an event with no file).
+ * analyzeChunkAccounting applies the same three conditions to the events it reads back.
+ */
+function resultFileOf(item) {
+  if (item === null || typeof item !== 'object' || !RESULT_TYPES.has(item.type)) return null;
+  const data = item.data;
+  if (data === null || typeof data !== 'object' || typeof data.file !== 'string') return null;
+  if (data.details && data.details.type === 'suite') return null;
+  return data.file;
+}
+
+/**
+ * One spelling for every ledger line: the real path. node reports a result's `file` already resolved
+ * through symlinks (macOS /var -> /private/var), while a zero-result child's only name is process.argv[1],
+ * as given, so without this the two kinds of line disagree on the same file. Best-effort: a path that cannot
+ * be resolved keeps its absolute spelling (analyzeChunkAccounting normalizes both sides again regardless).
+ */
+function canonicalPath(file) {
+  const absolute = path.resolve(file);
+  try {
+    return fs.realpathSync.native(absolute);
+  } catch {
+    return absolute;
+  }
+}
+
+function install(ledgerPath,serializerPrototype = DefaultSerializer.prototype) {
+  const counts = new Map(); // file -> leaf results handed to the serializer
+
+  const original = serializerPrototype.writeValue;
+  serializerPrototype.writeValue = function writeValueCountingResults(value) {
+    try {
+      const file = resultFileOf(value);
+      if (file !== null) counts.set(file, (counts.get(file) || 0) + 1);
+    } catch {
+      // Counting must never change what is serialized.
+    }
+    return Reflect.apply(original, this, arguments);
+  };
+
+  process.on('exit', () => {
+    try {
+      // A child that produced no result still records itself (count 0), so the runner can tell
+      // "the preload ran and saw nothing" from "the preload never ran".
+      if (counts.size === 0 && process.argv[1]) counts.set(process.argv[1], 0);
+      const lines = [];
+      for (const [file, count] of counts) {
+        lines.push(JSON.stringify({ type: 'registered', file: canonicalPath(file), count }));
+      }
+      fs.appendFileSync(ledgerPath, `${lines.join('\n')}\n`);
+    } catch {
+      // Best-effort: the ledger is a diagnostic channel and must never change
+      // the exit status of the test file it observes.
+    }
+  });
+}
+
+/**
+ * Make the test-file child's stdout (the pipe its results travel to the `node --test` parent on)
+ * blocking, so every frame it wrote has reached the OS when the child exits.
+ *
+ * Why: the runner passes `--test-force-exit`, and node forwards it to each file child. A forced
+ * child calls process.exit() as soon as its root test ends; its stdout is a non-blocking pipe on
+ * POSIX, so frames still queued in the Socket are discarded and the parent never receives those
+ * results (nodejs/node#64833; the fix proposed there, nodejs/node#64875, does exactly this with
+ * `_handle.setBlocking(true)` at force-exit). Measured on Node 24.18 / Linux with 20 files x 300
+ * tests, `--test-force-exit --test-concurrency=8`: 5039-5139 of 6000 results reported without
+ * this, 6000 of 6000 with it (and 6000 without force-exit). Doing it at load time rather than at
+ * exit means nothing is ever queued. The Windows hang guard that force-exit exists for (#1051)
+ * is untouched. Best-effort: a stdout with no pipe handle (a file, a TTY) is left alone.
+ */
+function blockChildStdout(stdout) {
+  try {
+    const target = stdout === undefined ? process.stdout : stdout;
+    const handle = target && target._handle;
+    if (handle && typeof handle.setBlocking === 'function') {
+      handle.setBlocking(true);
+      return true;
+    }
+  } catch {
+    // Never throw into the test file this observes.
+  }
+  return false;
+}
+
+// Inside a test-file child only (NODE_TEST_CONTEXT is set by `node --test` for those, not for the
+// runner parent): the blocking stdout needs no ledger path; the ledger count needs one.
+if (process.env.NODE_TEST_CONTEXT) {
+  blockChildStdout();
+  const ledgerPath = process.env.GSD_RUN_TESTS_LEDGER_FILE;
+  if (ledgerPath) install(ledgerPath);
+}
+
+module.exports = { install, resultFileOf, blockChildStdout };