Fix shared Web Fetch scoring and bind ten-method report evidence

.agents/skills/benchmark-report/SKILL.md

@@ -19,12 +19,13 @@ Public report CLIs default to English where supported and expose `--language en|
 
 1. Establish the decision being supported, the measured unit, cohort, versions, execution location, freshness policy, and intended deliverable. Inspect the actual dataset, runner, outputs, scorer, prior review evidence, and current report. Do not infer them from screenshots or prose.
 2. Freeze a machine-readable evaluation contract before expensive calls. It must identify the exact cases and methods, run count, timeout, concurrency policy, input and output contract, scoring provenance, reference state, and publication target when one exists. Existing compiled reference answers are frozen by default. Source metadata, capture dates, and absent newer acceptance-contract fields do not imply a live-reference task. Only tasks explicitly declaring `source_verification.reference_mode: live_exact` require live checks. For those tasks, fail closed: verify and freeze references immediately before calls, verify the same declared scope afterward, and certify the cohort only when both checks pass. An unchanged compiled acceptance-contract dataset is already the benchmark input: freeze its task and rule bytes for execution, without requiring historical source-qualification raw files or re-auditing cohort admission. Qualify sources when adding tasks or changing references/acceptance rules; do not turn an ordinary version rerun into dataset reconstruction. Partial or uncertified cohorts remain diagnostic evidence and cannot enter a release.
-3. Let each method use its strongest documented route to produce the required output. Equalize the task, inputs, output contract, call budget, and scoring. Do not require identical internals. Record material asymmetries such as local versus remote execution, cold versus warm start, bundled conversion, network transit, account tier, and billable time.
-4. Preserve evidence as an append-only chain: task definition, reference, provider call, raw artifact, normalized artifact, score, review, revision, and release. Keep provider output and call identity immutable. A presentation or scoring change must not masquerade as a new provider result.
+3. Let each method use its strongest documented route to produce the required output. Equalize the task, inputs, output contract, call budget, and scoring. Do not require identical internals. For fixed-native cohorts, declare the input media type independently of URL patterns and expected answers, freeze the class-level route and transport parameters before calls, and verify the retained request receipts at aggregation and release. Record material asymmetries such as local versus remote execution, cold versus warm start, bundled conversion, network transit, account tier, and billable time.
+4. Preserve evidence as an append-only chain: task definition, reference, provider call, raw artifact, normalized artifact, score, review, revision, and release. Keep provider output and call identity immutable. A presentation or scoring change must not masquerade as a new provider result. Declare the native answer fields and validate their structural projection before accepting a scored artifact. Adapters must not remove returned candidates by URL meaning or consult expected answers when selecting candidate fields. Apply privacy redaction only to separately identified public views after scoring, preserving the private originals and distinguishing original, scored and public-view hashes.
 5. Separate retrieval success, task acceptance, content quality, latency, cost, and freshness. A missing or partial reference makes the affected metric unavailable or scoped; it does not prove success or failure. Do not derive binary acceptance from an unrelated aggregate similarity score.
 6. For heterogeneous same-window latency evidence, preserve the full predeclared task-by-method matrix. Show all-call latency on that common denominator separately from successful-call latency on each method's conditional subset. Treat the all-method shared-pass intersection as a diagnostic, never as a post-outcome selector or ranking sample.
 7. At aggregate and publication boundaries, copy only explicit hash-bound evidence and public audit files. Reject symlinks and exclude worker trees, host paths, credentials, and internal run logs. Verify primary, auxiliary, delivered-file, and retained retry artifacts before release.
-8. Independent result review is opt-in: do not start it unless the user explicitly requests review. When requested without a broader scope, review machine failures only; review machine passes or every attempt only when the user explicitly asks for full review. Keep the initial verdict blind, but do not equate agreement with correctness: run the anomaly investigation in [audit and revision](references/audit-and-revision.md) for triggered results before calling the review complete. For a large independent review that is safely batchable, prefer resumable Codex CLI jobs with an exact model ID, read-only sandbox, strict output schema, non-overlapping packets, and bounded concurrency; persist each completed packet so a retry never repeats substantive review. Record the exact reviewed/total coverage, accepted unreviewed passes, reviewer, disagreements, adjudications, anomalies, and unresolved items. A review summary never replaces per-attempt evidence.
+   When a saved-input handoff needs host-metadata redaction, use the shared publication privacy owner. Preserve private originals and distinguish original from public file hashes; rebind dependent metadata references and disclose auxiliary debug-log projections. Never rewrite original scored answers, references or response payloads as a privacy repair. When a scored answer or native envelope itself contains private values, create a separately identified post-score public view through the shared privacy owner, verify its native-field projection and unchanged task verdict, and disclose that omitted original bytes cannot be reconstructed from the public copy. A public frozen presentation must explicitly bind its complete view inventory, verify displayed bytes against those views, keep acquisition gates strict, and distinguish preserved original diagnostics from what redacted text can reproduce. Verify unchanged verdicts, timings and review decisions, and rebuild from the public inputs. Keep regenerated package/report wrappers outside the original-input projection receipt so normal rebuilding does not invalidate it or create circular hashes.
+8. Independent result review is opt-in: do not start it unless the user explicitly requests review. When requested without a broader scope, review machine failures only; review machine passes or every attempt only when the user explicitly asks for full review. Label a review performed with known machine outcomes as an evidence audit, separately from independent blind-review coverage; a fresh review-origin tag alone does not establish blindness. Keep the initial verdict blind, but do not equate agreement with correctness: run the anomaly investigation in [audit and revision](references/audit-and-revision.md) for triggered results before calling the review complete. For a large independent review that is safely batchable, prefer resumable Codex CLI jobs with an exact model ID, read-only sandbox, strict output schema, non-overlapping packets, and bounded concurrency; persist each completed packet so a retry never repeats substantive review. Record the exact reviewed/total coverage, accepted unreviewed passes, reviewer, disagreements, adjudications, anomalies, and unresolved items. A review summary never replaces per-attempt evidence.
 For an explicitly requested full review, set `independent_result_review_scope: all` in the report profile and carry it into the assembled dataset comparison. The release boundary must reject failure-only review, missing attempts and accepted unreviewed passes. Reuse an existing content judgment only when the raw answer and acceptance contract are unchanged; verify each new call identity separately. For failed calls, inspect the current runtime evidence even when the answer is unchanged or empty: the same empty output can hide different failures. Changed evidence needs supplemental review, not another review of the unchanged population.
 
 9. Rerun only calls whose inputs, provider route/version, runtime contract, or source validity changed. Recompute downstream scores and reports without provider calls when only scoring, annotations, or presentation changed. Never retry merely to obtain a better result. A reference or verdict correction must identify the shared root cause and add executable positive/negative controls through the production scorer; bind those regression IDs in the revision receipt instead of patching stored verdicts.

evaluation/lexbench-webfetch.json

@@ -2,7 +2,7 @@
   "schema_version": 1,
   "kind": "external_benchmark_binding",
   "repository": "https://github.com/lexmount/lexbench-webfetch",
-  "revision": "6d8d72637bb96f4938b9a2f7c46d0c7dffd74945",
+  "revision": "c75efb365c7133d9f82f70fb1044f2ba88a36445",
   "ownership": {
     "benchmark": "Task populations, task model, acceptance scoring, file checks, compiler and frozen definitions",
     "harness": "Execution, reference lifecycle, evidence and reports"
@@ -78,7 +78,7 @@
     "evaluation/risk-control/remaining-70-rule-review.md": "301574dddcfddf6d30ed4d882cdd40b1ae859fdc95a0a02cf9868f55e3fd1079",
     "evaluation/risk-control/rule-authoring-guide.md": "74363ab502978179e5438a0b992bac7563090e41c91784e492636832dc1cd088",
     "evaluation/risk-control/source-inputs.json": "77c5a4b62c07d59ed57154a833e85f9e7a97bc739730f51cd4ca3a698edb9258",
-    "evaluation/scoring-regressions.json": "d519f86e327d9e2fd1c884670642eb31714435864e834592e2c7b65ef21ee655",
+    "evaluation/scoring-regressions.json": "159adaa11e5272adfe51555fddeba33283c5e5809a29a1d30bfc5ad911bcbe70",
     "evaluation/source-inputs.json": "c202abc4624811b938a837402bbaf0ecbe716e375fbca70567346d2edb3b1708"
   }
 }

harness/reporting/links.py

@@ -3,6 +3,7 @@
 
 import argparse
 import posixpath
+import re
 from html.parser import HTMLParser
 from pathlib import Path
 from urllib.parse import parse_qs, unquote, urlsplit
@@ -94,6 +95,12 @@ def parse_page(path: Path) -> Links:
         return parsed_pages[path]
 
     for page in pages:
+        # Frozen source HTML retains the original site's relative navigation.
+        # It is evidence, while report links pointing to it still need to exist.
+        relative = page.relative_to(root)
+        if "local-evidence" in relative.parts and not re.search(
+                r'<meta\b[^>]*name=[\'"]browser-eval-report[\'"]', page.read_text(encoding="utf-8", errors="replace")):
+            continue
         parser = parse_page(page)
         base_dir = root if (root / 'sample-manifest.json').is_file() else page.parent
         if parser.base_href: