[codex] Harden weekly metrics telemetry contracts

.github/scripts/__tests__/terminal-disposition-coverage.test.js

@@ -686,6 +686,35 @@ test('normalizes artifact selection reports without selected family counts', ()
   );
 });
 
+test('normalizes legacy artifact selection latest candidate maps', () => {
+  const summary = normalizeArtifactSelectionSummary({
+    status: 'pass',
+    candidate_family_counts: {
+      'review-thread-terminal-disposition': 2,
+    },
+    selected_family_counts: {},
+    latest_candidate_by_family: {
+      'review-thread-terminal-disposition': {
+        id: 44,
+        name: 'review-thread-terminal-disposition-latest',
+        created_at: '2026-04-25T12:00:00Z',
+      },
+    },
+    selected_artifacts: [],
+  });
+
+  const reviewThreadStatus = summary.terminal_priority_family_statuses.find(
+    (status) => status.family === 'review-thread-terminal-disposition'
+  );
+  assert.equal(reviewThreadStatus.status, 'available');
+  assert.equal(reviewThreadStatus.selected_artifact, null);
+  assert.deepEqual(reviewThreadStatus.latest_candidate, {
+    id: 44,
+    name: 'review-thread-terminal-disposition-latest',
+    created_at: '2026-04-25T12:00:00Z',
+  });
+});
+
 test('collects only terminal disposition ndjson files from metrics artifacts', () => {
   const dir = fs.mkdtempSync(path.join(os.tmpdir(), 'terminal-coverage-'));
   const terminalDir = path.join(dir, 'review-thread-terminal-disposition-123', 'agent-metrics');

.github/scripts/__tests__/weekly-metrics-artifacts.test.js

@@ -8,6 +8,7 @@ const {
   collectRepoArtifacts,
   formatArtifactTsv,
   formatSelectionMarkdown,
+  latestCandidateByFamily,
   missingPriorityFamilies,
   normalizeSelectionOptions,
   priorityFamilyStatuses,
@@ -102,6 +103,23 @@ test('selects only recent matching artifacts with a machine-readable report', ()
     'verifier-terminal-disposition',
     'bot-comment-auth-coverage-reusable',
   ]);
+  assert.deepEqual(
+    report.latest_candidate_by_family,
+    {
+      'bot-comment-auth-coverage-wrapper': {
+        id: 8,
+        name: 'bot-comment-auth-coverage-wrapper-77-2',
+        created_at: '2026-04-25T08:30:00Z',
+        updated_at: '2026-04-25T08:30:00Z',
+      },
+      'review-thread-terminal-disposition': {
+        id: 5,
+        name: 'review-thread-terminal-disposition-77',
+        created_at: '2026-04-25T11:00:00Z',
+        updated_at: '2026-04-25T11:00:00Z',
+      },
+    }
+  );
   assert.deepEqual(
     report.priority_family_statuses.map((family) => ({
       family: family.family,
@@ -250,6 +268,29 @@ test('reports priority telemetry families that are absent from the scan', () =>
   ]);
 });
 
+test('maps latest candidate artifacts by priority family', () => {
+  const candidates = [
+    artifact(1, 'review-thread-terminal-disposition-older', '2026-04-25T10:00:00Z'),
+    artifact(2, 'review-thread-terminal-disposition-newer', '2026-04-25T11:00:00Z'),
+    artifact(3, 'keepalive-metrics', '2026-04-25T12:00:00Z'),
+  ].map((raw) => ({
+    id: raw.id,
+    name: raw.name,
+    family: artifactFamily(raw.name),
+    created_at: raw.created_at,
+    updated_at: raw.updated_at,
+  }));
+
+  assert.deepEqual(latestCandidateByFamily(candidates), {
+    'review-thread-terminal-disposition': {
+      id: 2,
+      name: 'review-thread-terminal-disposition-newer',
+      created_at: '2026-04-25T11:00:00Z',
+      updated_at: '2026-04-25T11:00:00Z',
+    },
+  });
+});
+
 test('builds priority family statuses for available but unselected artifacts', () => {
   const candidates = [
     artifact(1, 'bot-comment-auth-coverage-wrapper-1', '2026-04-25T11:00:00Z'),

.github/scripts/terminal_disposition_coverage.js

@@ -440,6 +440,21 @@ function normalizeSelectionArtifact(artifact = {}) {
   return normalized;
 }
 
+function latestCandidateFromSelectionReport(report = {}, family = '') {
+  const candidateMaps = [
+    report.latest_candidate_by_family,
+    report.latest_candidates_by_family,
+  ];
+  for (const candidatesByFamily of candidateMaps) {
+    if (!candidatesByFamily || typeof candidatesByFamily !== 'object' || Array.isArray(candidatesByFamily)) {
+      continue;
+    }
+    const candidate = normalizeSelectionArtifact(candidatesByFamily[family]);
+    if (candidate) return candidate;
+  }
+  return null;
+}
+
 function normalizeTerminalPriorityFamilyStatuses(report = {}, selectedArtifacts = []) {
   const rawStatuses = Array.isArray(report.priority_family_statuses)
     ? report.priority_family_statuses
@@ -466,12 +481,14 @@ function normalizeTerminalPriorityFamilyStatuses(report = {}, selectedArtifacts
     const selectedCount = Number(report.selected_family_counts?.[family]) ||
       selectedArtifacts.filter((artifact) => artifact.family === family).length;
     const selectedArtifact = selectedArtifacts.find((artifact) => artifact.family === family) || null;
+    const latestCandidate = latestCandidateFromSelectionReport(report, family) ||
+      (selectedArtifact ? normalizeSelectionArtifact(selectedArtifact) : null);
     byFamily.set(family, {
       family,
       status: selectedCount > 0 ? 'selected' : (candidateCount > 0 ? 'available' : 'missing'),
       candidate_count: candidateCount,
       selected_count: selectedCount,
-      latest_candidate: selectedArtifact ? normalizeSelectionArtifact(selectedArtifact) : null,
+      latest_candidate: latestCandidate,
       selected_artifact: selectedArtifact ? normalizeSelectionArtifact(selectedArtifact) : null,
     });
   }

.github/scripts/weekly_metrics_artifacts.js

@@ -166,6 +166,41 @@ function priorityFamilyStatuses({
   });
 }
 
+function latestCandidateByFamily(candidates = []) {
+  const latestByFamily = new Map();
+  for (const candidate of candidates) {
+    const family = cleanString(candidate.family);
+    if (!PRIORITY_METRICS_FAMILIES.includes(family)) continue;
+    const existing = latestByFamily.get(family);
+    const candidateTimestamp = Number(candidate.timestamp_ms) || artifactTimestampMs(candidate);
+    const existingTimestamp = existing ? Number(existing.timestamp_ms) || artifactTimestampMs(existing) : -1;
+    const candidateId = Number(candidate.id) || 0;
+    const existingId = existing ? Number(existing.id) || 0 : -1;
+    if (
+      !existing ||
+      candidateTimestamp > existingTimestamp ||
+      (candidateTimestamp === existingTimestamp && candidateId > existingId)
+    ) {
+      latestByFamily.set(family, candidate);
+    }
+  }
+  const entries = [];
+  for (const family of PRIORITY_METRICS_FAMILIES) {
+    const latestCandidate = latestByFamily.get(family);
+    if (!latestCandidate) continue;
+    entries.push([
+      family,
+      {
+        id: latestCandidate.id,
+        name: latestCandidate.name,
+        created_at: latestCandidate.created_at,
+        updated_at: latestCandidate.updated_at,
+      },
+    ]);
+  }
+  return Object.fromEntries(entries);
+}
+
 function selectMetricsArtifacts(artifacts = [], options = {}) {
   const config = normalizeSelectionOptions(options);
   const stats = {
@@ -261,6 +296,7 @@ function selectMetricsArtifacts(artifacts = [], options = {}) {
     ...stats,
     candidate_family_counts: sortedCountObject(candidateFamilyCounts),
     selected_family_counts: sortedCountObject(familyCounts),
+    latest_candidate_by_family: latestCandidateByFamily(candidates),
     missing_priority_families: missingPriorityFamilies(candidateFamilyCounts),
     priority_family_statuses: priorityFamilyStatuses({
       candidates,
@@ -302,6 +338,7 @@ function buildSelectionErrorReport(options = {}, error = {}) {
     ignored_total_limit_count: 0,
     candidate_family_counts: {},
     selected_family_counts: {},
+    latest_candidate_by_family: {},
     missing_priority_families: [...PRIORITY_METRICS_FAMILIES],
     priority_family_statuses: priorityFamilyStatuses(),
     selected_artifacts: [],
@@ -489,6 +526,7 @@ module.exports = {
   collectRepoArtifacts,
   formatArtifactTsv,
   formatSelectionMarkdown,
+  latestCandidateByFamily,
   missingPriorityFamilies,
   normalizeSelectionOptions,
   priorityFamilyStatuses,

scripts/aggregate_agent_metrics.py

@@ -44,6 +44,8 @@
     ),
 )
 _MAX_PARSE_ERROR_ROWS = 25
+_MAX_LEGACY_JSON_FALLBACK_LINES = 5000
+_MAX_LEGACY_JSON_FALLBACK_BYTES = 1024 * 1024
 
 
 @dataclass(frozen=True)
@@ -180,13 +182,25 @@ def _read_ndjson(files: Iterable[Path]) -> tuple[list[dict[str, Any]], list[Pars
         file_entries: list[dict[str, Any]] = []
         file_errors: list[ParseErrorDetail] = []
         raw_lines_for_fallback: list[str] = []
+        raw_fallback_bytes = 0
+        raw_fallback_truncated = False
         with handle:
             for line_number, line in enumerate(handle, start=1):
                 raw = line.strip()
                 if not raw:
                     continue
-                if not file_entries:
-                    raw_lines_for_fallback.append(raw)
+                if not file_entries and not raw_fallback_truncated:
+                    raw_bytes = len(raw.encode("utf-8")) + 1
+                    fallback_within_limit = (
+                        len(raw_lines_for_fallback) < _MAX_LEGACY_JSON_FALLBACK_LINES
+                        and raw_fallback_bytes + raw_bytes <= _MAX_LEGACY_JSON_FALLBACK_BYTES
+                    )
+                    if fallback_within_limit:
+                        raw_fallback_bytes += raw_bytes
+                        raw_lines_for_fallback.append(raw)
+                    else:
+                        raw_fallback_truncated = True
+                        raw_lines_for_fallback = []
                 try:
                     parsed = json.loads(raw)
                 except json.JSONDecodeError:
@@ -197,7 +211,14 @@ def _read_ndjson(files: Iterable[Path]) -> tuple[list[dict[str, Any]], list[Pars
                     raw_lines_for_fallback = []
                 else:
                     file_errors.append(_parse_error_detail(path, line_number, "non-object-json"))
-        if file_errors and not file_entries and raw_lines_for_fallback:
+        if file_errors and not file_entries and raw_fallback_truncated:
+            file_errors.append(_parse_error_detail(path, None, "legacy-json-fallback-buffer-limit"))
+        if (
+            file_errors
+            and not file_entries
+            and raw_lines_for_fallback
+            and not raw_fallback_truncated
+        ):
             try:
                 parsed_file = json.loads("\n".join(raw_lines_for_fallback))
             except json.JSONDecodeError:

scripts/repo_review_evaluator.py

@@ -444,11 +444,7 @@ def gap_label(severity: str) -> str:
 def build_review_execution(state: dict[str, Any]) -> dict[str, Any]:
     repo_path = Path(state["local_path"])
     tracked_files = tracked_repo_files(repo_path)
-    implementation_files = [
-        path
-        for path in tracked_files
-        if is_implementation_file(path)
-    ]
+    implementation_files = [path for path in tracked_files if is_implementation_file(path)]
     test_files = [path for path in tracked_files if is_test_file(path)]
     implementation_scan_files = implementation_files[:REVIEW_SCAN_FILE_LIMIT]
     test_scan_files = test_files[:REVIEW_SCAN_FILE_LIMIT]
@@ -499,8 +495,7 @@ def build_review_execution(state: dict[str, Any]) -> dict[str, Any]:
         :12
     ]
     design_headings = {
-        rel_path: markdown_headings(repo_path, rel_path)
-        for rel_path in state["design_files"][:6]
+        rel_path: markdown_headings(repo_path, rel_path) for rel_path in state["design_files"][:6]
     }
     domain_hits = keyword_file_hits(
         repo_path,
@@ -521,12 +516,12 @@ def build_review_execution(state: dict[str, Any]) -> dict[str, Any]:
 
     if not state["design_files"]:
         design_severity = "blocks testing"
-        design_finding = "No tracked design sources were found; the design contract is not reviewable."
-    else:
-        design_severity = "none"
         design_finding = (
-            f"Collected {state['design_source_count']} design sources and registry anchor for comparison."
+            "No tracked design sources were found; the design contract is not reviewable."
         )
+    else:
+        design_severity = "none"
+        design_finding = f"Collected {state['design_source_count']} design sources and registry anchor for comparison."
     dimensions.append(
         {
             "id": "design_contract",
@@ -580,14 +575,10 @@ def build_review_execution(state: dict[str, Any]) -> dict[str, Any]:
         test_finding = "No tests or CI workflow files were detected."
     elif not smoke_test_files:
         test_severity = "material"
-        test_finding = (
-            "Tests or workflows exist, but the automated pass did not find smoke/e2e/live readiness markers."
-        )
+        test_finding = "Tests or workflows exist, but the automated pass did not find smoke/e2e/live readiness markers."
     else:
         test_severity = "needs human decision"
-        test_finding = (
-            "Test and smoke/integration markers exist; review must verify they prove the intended user journey."
-        )
+        test_finding = "Test and smoke/integration markers exist; review must verify they prove the intended user journey."
     dimensions.append(
         {
             "id": "test_and_live_readiness",
@@ -628,14 +619,10 @@ def build_review_execution(state: dict[str, Any]) -> dict[str, Any]:
 
     if state["issue_draft_count"] or state["archive_candidate_count"]:
         issue_severity = "needs human decision"
-        issue_finding = (
-            "Draft inputs exist; approve only after checking them against the executed review evidence."
-        )
+        issue_finding = "Draft inputs exist; approve only after checking them against the executed review evidence."
     elif state["remote_open_issue_count"]:
         issue_severity = "needs human decision"
-        issue_finding = (
-            "No local/archive candidates are queued, but remote open issues need reconciliation before drafting more."
-        )
+        issue_finding = "No local/archive candidates are queued, but remote open issues need reconciliation before drafting more."
     else:
         issue_severity = "needs human decision"
         issue_finding = (
@@ -1367,9 +1354,7 @@ def write_repo_artifacts(output_dir: Path, state: dict[str, Any], max_drafts: in
             "",
         ]
     )
-    (repo_dir / "review-execution.md").write_text(
-        "\n".join(execution_lines), encoding="utf-8"
-    )
+    (repo_dir / "review-execution.md").write_text("\n".join(execution_lines), encoding="utf-8")
 
     draft_lines = [
         f"# Issue Drafts: {state['repo']}",
@@ -1432,19 +1417,15 @@ def write_packet(output_dir: Path, states: list[dict[str, Any]], generated_on: s
     ignored = [state for state in states if state["status"] == "ignored"]
     review_pending = [state for state in active if state["review_status"] == PENDING_REVIEW_STATUS]
     blocked = [
-        state
-        for state in active
-        if str(state["review_execution"]["status"]).startswith("blocked")
+        state for state in active if str(state["review_execution"]["status"]).startswith("blocked")
     ]
     issue_candidate_repos = [
         state for state in active if state["issue_queue_status"] == "draft candidates present"
     ]
     executed_reviews = [
         state for state in active if state["review_execution"]["status"] == "executed"
     ]
-    automated_gap_repos = [
-        state for state in active if state["review_execution"]["gap_count"] > 0
-    ]
+    automated_gap_repos = [state for state in active if state["review_execution"]["gap_count"] > 0]
 
     lines = [
         f"# Weekly Design Review Decision Packet - {generated_on}",
@@ -1481,7 +1462,9 @@ def write_packet(output_dir: Path, states: list[dict[str, Any]], generated_on: s
                 "approve/edit/defer issue drafts."
             )
         elif str(state["review_execution"]["status"]).startswith("blocked"):
-            human_action = "resolve the blocker, rerun review execution, then queue the human decision."
+            human_action = (
+                "resolve the blocker, rerun review execution, then queue the human decision."
+            )
         else:
             human_action = "conduct the standardized review, then approve/edit/defer issue drafts."
         lines.extend(