@opengsd/gsd-core 1.14.0 → 1.16.0

package/scripts/gen-plugin-skills.cjs

@@ -17,6 +17,21 @@
  * installer uses, producing a build-generated skills/ dir that ships in the
  * npm package and serves plugin-only installs.
  *
+ * #4995: commands/gsd/*.md still instruct bare command-position `gsd-tools
+ * <verb>` calls in 4 files (workstreams.md, quick.md, review-backlog.md,
+ * config.md), which fail with "command not found" on a shim-only install.
+ * That source can't be normalized to `gsd_run` directly the way #2751 did for
+ * agents/*.md and gsd-core/workflows/*.md: tests/gsd-tools-path-refs.test.cjs
+ * (#1766) pins commands/gsd/workstreams.md to the literal string
+ * `gsd-tools query workstream.list`, and bringing commands/ under the #2751
+ * guard was tried and reverted for unrelated load-bearing reasons (11ad7881ed
+ * / d286e54163). So this generator rewrites bare gsd-tools -> gsd_run in its
+ * OWN output — the same shape of fix #725 already established for the Codex
+ * conversion pipeline (rewrite the generated artifact, not the shared
+ * source) — reusing the exact `gsd_run` resolver preamble #2751 established
+ * (scripts/sync-runtime-launcher.cjs's transformFile) rather than inventing a
+ * new resolution mechanism.
+ *
  * Depends on: gsd-core/bin/lib/runtime-artifact-conversion.cjs (compiled from
  * src/runtime-artifact-conversion.cts by `npm run build:lib`). Must run AFTER
  * build:lib in the build chain.
@@ -25,6 +40,7 @@
 const fs = require('node:fs');
 const path = require('node:path');
 const { ExitError, runMain } = require('./lib/cli-exit.cjs');
+const { transformFile, loadPreamble } = require('./sync-runtime-launcher.cjs');
 
 const ROOT = path.resolve(__dirname, '..');
 const COMMANDS_DIR = path.join(ROOT, 'commands', 'gsd');
@@ -33,6 +49,73 @@ const CONVERSION_MODULE = path.join(ROOT, 'gsd-core', 'bin', 'lib', 'runtime-art
 const PREFIX = 'gsd-';
 const RUNTIME = 'claude';
 
+// #4995: the 4 commands/gsd/*.md stems (census taken by the issue) that carry
+// command-position bare `gsd-tools <verb>` calls. Deliberately an explicit
+// allowlist, NOT a blanket regex applied to every generated skill: several
+// other skills (e.g. gsd-graphify) merely NAME `gsd-tools`/`gsd_run`
+// descriptively (prose, "DO NOT use ..." warnings) in ways a command-position
+// regex cannot safely tell apart from a real instruction without per-site
+// human review -- exactly the ambiguity tests/no-bare-gsd-tools-command-
+// position.test.cjs's PROSE_ALLOWLIST exists to resolve by hand for #2751/
+// #3809. gsd-graphify also carries 5 deliberately-separate per-block
+// preambles (tests/graphify-visualization.test.cjs extracts and runs each
+// fenced Step-3 block standalone); running every skill through
+// transformFile's "collapse to one preamble per file" behavior would
+// reproduce the exact regression d286e54163 already found and reverted for
+// commands/gsd/graphify.md. Restricting the rewrite to the 4 files the
+// census in #4995 actually names avoids all of that.
+const BARE_GSD_TOOLS_STEMS = new Set(['workstreams', 'quick', 'review-backlog', 'config']);
+
+// #4995: command-position bare `gsd-tools` -> `gsd_run` token swap. Mirrors 3
+// of the 4 command-position shapes rewriteBareGsdToolsCommandsForCodex
+// (src/runtime-artifact-conversion.cts) already covers for the Codex
+// pipeline: start-of-line, inside `$( ... )`, and immediately after a
+// backtick (inline "Run: `gsd-tools verb args`" prose). Deliberately DROPS
+// that function's 4th shape -- a bare, unescaped `|` counted as a shell
+// pipe separator -- because a Markdown table row (`| col | gsd-tools verb
+// arg |`) also starts a cell with `| gsd-tools`, and none of the 11 sites in
+// #4995's census need it: config.md's own routing TABLE names `gsd-tools
+// query config-set-model-profile` descriptively in a `|`-delimited cell one
+// row above the real operative site, and rewriting that cell too would be a
+// change nobody asked for. (`&&`/`;` are kept: two-character or dedicated
+// separators, not also a Markdown table delimiter.) Only applied to
+// BARE_GSD_TOOLS_STEMS files (see above). Within those 4 files this narrower
+// pattern still matches one site beyond the issue's 11-site census --
+// gsd-quick/SKILL.md's <security_notes> line ("Status fields read via
+// `gsd-tools query frontmatter.get`") is a backtick-wrapped DESCRIPTIVE
+// mention, not a `Run:` instruction, but it has the identical command-
+// position shape and gets swapped too. That's harmless (the sentence reads
+// the same with `gsd_run` in it) and left as-is rather than special-cased,
+// but the true count this generator fixes is 12 operative-shaped sites, not
+// 11 -- the 12th just wasn't literal-command "operative" in the issue's own
+// sense.
+function rewriteBareGsdToolsCommandsToGsdRun(content) {
+  return content
+    .replace(/(^[ \t]*)gsd-tools(?=\s)/gm, '$1gsd_run')
+    .replace(/(\$\(\s*)gsd-tools(?=\s)/g, '$1gsd_run')
+    .replace(/(`\s*)gsd-tools(?=\s)/g, '$1gsd_run')
+    .replace(/((?:&&|;)\s*)gsd-tools(?=\s)/g, '$1gsd_run');
+}
+
+// #4995: after the token swap, any fenced bash/sh/shell block that now calls
+// `gsd_run` needs the resolver preamble that defines it -- reuse
+// sync-runtime-launcher.cjs's transformFile verbatim (the same mechanism
+// applied to agents/*.md and gsd-core/workflows/*.md) rather than
+// reimplementing preamble placement here. Inline (non-fenced) "Run: `gsd_run
+// verb args`" prose has no fenced block to anchor a preamble to; per the
+// precedent already set for this exact shape (commands/gsd/workstreams.md
+// and config.md in 11ad7881ed), it is left as a bare `gsd_run` token --
+// `gsd_run` is itself a shipped npm bin name, so it resolves identically to
+// how the pre-fix `gsd-tools` form resolved once PATH carries it (e.g. via
+// the CLAUDE_ENV_FILE export in an earlier-run preamble in the same
+// session), without regressing anything that worked before.
+function rewriteSkillContent(content) {
+  const swapped = rewriteBareGsdToolsCommandsToGsdRun(content);
+  const preamble = loadPreamble();
+  const withPreamble = transformFile(swapped, preamble);
+  return withPreamble === null ? swapped : withPreamble;
+}
+
 function generateSkills(conversion) {
   const cmdNames = conversion.readGsdCommandNames();
   const files = fs.readdirSync(COMMANDS_DIR).filter(f => f.endsWith('.md'));
@@ -41,7 +124,10 @@ function generateSkills(conversion) {
     const stem = file.slice(0, -3);
     const skillName = PREFIX + stem;
     const src = fs.readFileSync(path.join(COMMANDS_DIR, file), 'utf8');
-    const converted = conversion.convertClaudeCommandToClaudeSkill(src, skillName, RUNTIME, cmdNames, true);
+    let converted = conversion.convertClaudeCommandToClaudeSkill(src, skillName, RUNTIME, cmdNames, true);
+    if (BARE_GSD_TOOLS_STEMS.has(stem)) {
+      converted = rewriteSkillContent(converted);
+    }
     results.push({ skillName, content: converted });
   }
   return results;

package/scripts/gen-research-agents.cjs

@@ -25,6 +25,9 @@ const path = require('node:path');
 
 const { PROFILES } = require('./research-profiles.cjs');
 const { ExitError, runMain } = require('./lib/cli-exit.cjs');
+// The one frontmatter fence owner (src/frontmatter-fence.cts, built by build:lib).
+// This dev-time script runs after the build (manually, and from the tests).
+const { locateFrontmatterFence } = require('../gsd-core/bin/lib/frontmatter-fence.cjs');
 
 const ROOT = path.resolve(__dirname, '..');
 const AGENTS_DIR = path.join(ROOT, 'agents');
@@ -41,12 +44,13 @@ const AGENTS_DIR = path.join(ROOT, 'agents');
  * We read the CURRENT commented-hooks block from the agent file and preserve it
  * byte-for-byte; only name/description/tools/color are regenerated.
  */
-function buildFrontmatter(profile, existingFrontmatter) {
-  // Extract the commented hooks section from the existing frontmatter, if any.
-  // The hooks block starts at `# hooks:` and runs to (but not including) the
-  // closing `---`. In the committed files there is NO blank line between
-  // `color:` and `# hooks:`, so we append it directly after the color line's `\n`.
-  const hooksMatch = existingFrontmatter.match(/(# hooks:[\s\S]*?)(?=\n---)/);
+function buildFrontmatter(profile, existingRegion) {
+  // Extract the commented hooks section from the existing frontmatter region
+  // (the YAML text between the fences), if any. The hooks block starts at
+  // `# hooks:` and runs to the end of the region — the closing `---` follows it.
+  // In the committed files there is NO blank line between `color:` and
+  // `# hooks:`, so we append it directly after the color line's `\n`.
+  const hooksMatch = existingRegion.match(/(# hooks:[\s\S]*)$/);
   // hooksSuffix: if present, the block followed by a newline so `---` is on its own line;
   // if absent, empty string (the closing `---` follows directly after color's `\n`).
   const hooksSuffix = hooksMatch ? hooksMatch[1] + '\n' : '';
@@ -67,37 +71,26 @@ function buildFrontmatter(profile, existingFrontmatter) {
 /**
  * Parse a .md file and return { frontmatterRaw, body, frontmatterFields }.
  *
- * frontmatterRaw: the raw text between the first and second `---` delimiters (exclusive)
- * body: everything after the closing `---\n`
+ * frontmatterRaw: the YAML text between the opening and closing fences (exclusive)
+ * fullFrontmatter: the block from the opening fence through the closing fence line
+ * body: everything after the closing fence line's line ending
  * frontmatterFields: { name, description, color, tools }
+ *
+ * The block is the one the one fence owner finds (found while implementing #5105):
+ * it opens at byte 0 (after a BOM), exactly as the runtime that loads the agent reads it.
  */
 function parseAgentFile(filePath) {
   const raw = fs.readFileSync(filePath, 'utf8');
 
-  // The frontmatter is between the first `---` line and the next `---` line.
-  const lines = raw.split('\n');
-  let start = -1;
-  let end = -1;
-  for (let i = 0; i < lines.length; i++) {
-    if (lines[i].trim() === '---') {
-      if (start === -1) {
-        start = i;
-      } else {
-        end = i;
-        break;
-      }
-    }
-  }
-
-  if (start === -1 || end === -1) {
+  const fence = locateFrontmatterFence(raw);
+  if (!fence || !fence.closed) {
     throw new Error('No valid frontmatter delimiters found in ' + filePath);
   }
 
-  const frontmatterLines = lines.slice(start + 1, end);
-  const frontmatterRaw = frontmatterLines.join('\n');
-  // body includes the closing `---` line and everything after
-  const fullFrontmatter = lines.slice(start, end + 1).join('\n');
-  const body = lines.slice(end + 1).join('\n');
+  const frontmatterRaw = raw.slice(fence.openEnd, fence.bodyEnd);
+  const frontmatterLines = frontmatterRaw.split(/\r?\n/);
+  const fullFrontmatter = raw.slice(fence.bom.length, fence.closingFenceEnd);
+  const body = raw.slice(fence.closingFenceEnd).replace(/^\r?\n/, '');
 
   const fields = {};
   // Parse simple key: value pairs (not nested YAML, no multi-line values here)
@@ -214,9 +207,9 @@ function runCheck() {
  */
 function writeAgent(profile) {
   const agentPath = path.join(AGENTS_DIR, profile.name + '.md');
-  const { fullFrontmatter, body } = parseAgentFile(agentPath);
+  const { frontmatterRaw, body } = parseAgentFile(agentPath);
 
-  const newFrontmatter = buildFrontmatter(profile, fullFrontmatter);
+  const newFrontmatter = buildFrontmatter(profile, frontmatterRaw);
   const newContent = newFrontmatter + '\n' + body;
 
   fs.writeFileSync(agentPath, newContent, 'utf8');

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 +33,8 @@ 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",
   "tests/chunked-planning-parallel.test.cjs",
@@ -43,7 +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",
@@ -52,6 +55,7 @@ module.exports = {
   "tests/config-loader.test.cjs",
   "tests/config-schema.property.test.cjs",
   "tests/config.test.cjs",
+  "tests/configured-entrypoint-validation.test.cjs",
   "tests/contributor-standards.test.cjs",
   "tests/core-utils.test.cjs",
   "tests/cursor-subagent-isolation.test.cjs",
@@ -59,6 +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",
@@ -73,16 +78,24 @@ 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",
   "tests/gsd-tools-path-refs.test.cjs",
   "tests/gsd-validate-commit-crash-policy.test.cjs",
+  "tests/gsd-validate-commit-sigpipe.test.cjs",
   "tests/gsd-write-guard.test.cjs",
   "tests/health-diagnostic-rules/worktree-health.test.cjs",
   "tests/health-diagnostic.test.cjs",
@@ -92,7 +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",
@@ -111,8 +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",
@@ -125,6 +140,7 @@ module.exports = {
   "tests/no-posix-mode-bit-assert.rule.test.cjs",
   "tests/no-private-binary-resolution.rule.test.cjs",
   "tests/no-unguarded-nonportable-exec.rule.test.cjs",
+  "tests/nsegment-phase-grammar.test.cjs",
   "tests/observability/event.test.cjs",
   "tests/onboard-command.test.cjs",
   "tests/opencode-plugin-adapter.test.cjs",
@@ -136,8 +152,11 @@ 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",
   "tests/plan-phase-stall-detection.test.cjs",
   "tests/plan-review-convergence.test.cjs",
   "tests/planning-inspect.test.cjs",
@@ -154,6 +173,8 @@ module.exports = {
   "tests/process-seam.test.cjs",
   "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",
@@ -173,21 +194,31 @@ module.exports = {
   "tests/roadmap-mode-field.test.cjs",
   "tests/roadmap-parser.test.cjs",
   "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",
   "tests/runtime-launcher-parity.test.cjs",
+  "tests/runtime-name-policy.test.cjs",
   "tests/security.test.cjs",
   "tests/settings-jsonc.test.cjs",
   "tests/shared-hooks-dir-resolution.test.cjs",
   "tests/shell-command-projection-dispatch.test.cjs",
+  "tests/shipped-reference-cites.test.cjs",
   "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",
   "tests/unreachable-guard-drift.test.cjs",
@@ -195,6 +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

@@ -16,8 +16,9 @@
 // Durability, not `--test-reporter-destination` (#3889 root cause): a
 // reporter that YIELDS strings has them piped by Node into a
 // `fs.WriteStream` targeting the destination path, and that stream buffers.
-// The parent's `execFileSync` timeout SIGKILLs the child on a hang, and
-// SIGKILL is uncatchable and gives the process zero chance to flush — so a
+// The parent's per-chunk timeout (run-tests.cjs's runChunk) kills the child
+// on a hang, and a hard kill (TerminateProcess on Windows, SIGKILL on POSIX)
+// is uncatchable and gives the process zero chance to flush — so a
 // yield-based reporter can lose every event still sitting in the stream's
 // buffer, which is exactly the case this feature exists to diagnose (proven
 // live: a chunk killed at 2006ms produced a `killed after 2006ms` line from
@@ -49,8 +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
@@ -93,13 +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 = `${JSON.stringify({
         type: event.type,
         file,
         name,
         nesting,
         testNumber,
+        kind,
         ts: Date.now(),
       })}\n`;
       try {
@@ -108,6 +116,24 @@ module.exports = async function ndjsonEventReporter(source) {
         // Best-effort: a write failure here (e.g. the events dir vanished)
         // must never crash the test run this reporter is only observing.
       }
+    } else if (event.type === 'test:summary' && typeof (event.data || {}).file === 'string') {
+      // #5071: node:test emits one `test:summary` per test FILE (its
+      // `duration_ms` is that file's wall-clock) plus a run-level one with no
+      // `file`. Only the per-file ones are recorded: run-tests.cjs exports
+      // them, when asked, as the raw input for a per-platform timing table
+      // (scripts/gen-test-timings.cjs --platform). The run-level summary is
+      // not a file measurement and is skipped here.
+      const line = `${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