@lcv-ideas-software/cross-review 4.2.5 → 4.3.1

package/CHANGELOG.md

@@ -7,6 +7,54 @@ standard `v00.00.00`; npm package versions remain SemVer.
 
 ## [Unreleased]
 
+## [v04.03.01] — 2026-06-05
+
+**Patch — provider skip classification hotfix.** This release follows up on a
+real hard-gate incident where Claude/Anthropic was skipped after provider
+overload. The immediate provider cause was Anthropic `overloaded_error`; the
+runtime issue was that any `provider_error`, including non-retryable provider
+400-style failures, could be treated as skippable.
+
+### Changed
+
+- `provider_error` is now skippable only when classified as retryable, so
+  non-retryable provider payload/schema rejections block convergence instead of
+  being silently removed from the panel.
+- Anthropic `overloaded_error` without preserved HTTP status text is now treated
+  as retryable, matching HTTP 529 overload behavior.
+- `session.peer_skipped_unavailable` events now include retryability, recovery
+  hint, and a redacted provider error preview in event data.
+
+## [v04.03.00] — 2026-06-05
+
+**Minor — P1/P2/P3 audit follow-up.** This release closes the first concrete
+items from the post-v4.2.5 runtime/session audit: unresolved evidence is harder
+to miss at finalization time, fixture-level regressions can be evaluated
+offline, and operators get a read-only peer reliability report without changing
+peer selection.
+
+### Added
+
+- Added `session_peer_reliability_report`, a read-only MCP tool that aggregates
+  per-peer parser warnings, decision quality, rejected/provider failures,
+  evidence checklist dispositions, fabrication-related events, latency and
+  cost.
+- Added `npm run eval:fixtures`, an offline fixture harness for truthfulness
+  preflight, parser diagnostics and report rendering contracts. It does not
+  start provider sessions or call reviewers.
+- `session_report` now includes an **Unresolved Evidence Disposition** section
+  when checklist items remain `open` or `not_resurfaced`.
+
+### Changed
+
+- Automatic convergence with unresolved checklist items now finalizes with
+  `unanimous_ready_with_unresolved_evidence` or
+  `recovered_unanimity_with_unresolved_evidence` instead of a plain success
+  reason.
+- Finalization now emits `session.evidence_checklist_unresolved_on_finalize`
+  with unresolved counts and item summaries when a session closes while
+  evidence asks are still open or only inferred as not resurfaced.
+
 ## [v04.02.05] — 2026-06-05
 
 **Patch — session audit hardening.** This release closes follow-ups from the

package/README.md

@@ -24,7 +24,7 @@ npm install -g @lcv-ideas-software/cross-review
 npm install -g @lcv-ideas-software/cross-review --registry=https://npm.pkg.github.com
 ```
 
-**Status.** Stable. Current release: **v04.02.05** (npm package `4.2.5`). See [CHANGELOG.md](./CHANGELOG.md) for the full release history.
+**Status.** Stable. Current release: **v04.03.01** (npm package `4.3.1`). See [CHANGELOG.md](./CHANGELOG.md) for the full release history.
 
 > **Project renamed 2026-05-15.** This project was previously published as
 > [`@lcv-ideas-software/cross-review-v2`](https://www.npmjs.com/package/@lcv-ideas-software/cross-review-v2)
@@ -38,6 +38,8 @@ The version history at a glance:
 
 | Release              | Scope                                                                                                                                                                                                              |
 | -------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ |
+| **`v04.03.01`**      | Patch — tighten skip-peer classification so non-retryable provider errors block, while Anthropic overload events remain retryable and better surfaced in skip diagnostics.                                         |
+| **`v04.03.00`**      | Minor — P1/P2/P3 follow-up with unresolved-evidence close-out visibility, an offline fixture eval harness, and a read-only peer reliability report.                                                                |
 | **`v04.02.05`**      | Patch — harden session auditability with terminal events, cost split reporting, `not_resurfaced` visibility, and relator provenance checks for session IDs/GitHub URLs.                                            |
 | **`v04.02.04`**      | Patch — harden truthfulness preflight auditability, add a read-only preflight retest tool, and reduce false parser warnings for attached/log evidence.                                                             |
 | **`v04.02.03`**      | Patch — promote the Gemini canonical default to `gemini-3.1-pro-preview` and refresh the active local Gemini rate card.                                                                                            |
@@ -211,6 +213,7 @@ these environment variables before running real sessions (example):
 - `session_metrics`
 - `session_doctor`
 - `session_report`
+- `session_peer_reliability_report`
 - `session_check_convergence`
 - `session_truthfulness_preflight_check`
 - `session_attach_evidence`
@@ -228,7 +231,12 @@ these environment variables before running real sessions (example):
 lack terminal events, and reports peer-call cost separately from generation
 artifact cost. `session_report` uses the same split and calls out
 `not_resurfaced` evidence checklist items as inference-only, not proof that the
-requested evidence was satisfied.
+requested evidence was satisfied. If a session otherwise reaches unanimity with
+open or `not_resurfaced` checklist items, finalization records an
+`*_with_unresolved_evidence` outcome reason and emits a durable unresolved
+evidence event. `session_peer_reliability_report` is read-only and aggregates
+per-peer parser warnings, evidence ask status, provider failures, cost and
+latency.
 
 ## Repository conventions
 

package/dist/scripts/eval-fixtures.d.ts

@@ -0,0 +1,54 @@
+export declare const truthfulnessCases: readonly [{
+    readonly name: "current runtime contradiction is blocked";
+    readonly input: {
+        readonly task: "The current cross-review runtime is 4.2.4.";
+        readonly runtimeFacts: {
+            readonly runtime_version: "4.2.5";
+            readonly release_date: "2026-06-05";
+        };
+        readonly attachmentsPresent: false;
+    };
+    readonly expectPass: false;
+    readonly expectIssueClass: "runtime_contradiction";
+}, {
+    readonly name: "matching current runtime facts pass";
+    readonly input: {
+        readonly task: "server_info shows current cross-review runtime 4.2.5.";
+        readonly runtimeFacts: {
+            readonly runtime_version: "4.2.5";
+            readonly release_date: "2026-06-05";
+        };
+        readonly attachmentsPresent: false;
+    };
+    readonly expectPass: true;
+}, {
+    readonly name: "historical timing claim needs snapshot evidence";
+    readonly input: {
+        readonly task: "When the audit began, cross-review was running 4.2.4.";
+        readonly runtimeFacts: {
+            readonly runtime_version: "4.2.5";
+            readonly release_date: "2026-06-05";
+        };
+        readonly attachmentsPresent: false;
+    };
+    readonly expectPass: false;
+    readonly expectIssueClass: "unsupported_historical_claim";
+}];
+export declare const parserCases: readonly [{
+    readonly name: "verified with empty evidence gets empty-evidence warning";
+    readonly text: string;
+    readonly expectStatus: "READY";
+    readonly expectWarning: "verified_without_evidence_sources";
+}, {
+    readonly name: "verified with attached evidence path is concrete";
+    readonly text: string;
+    readonly expectStatus: "READY";
+    readonly absentWarning: "verified_without_evidence_sources";
+}];
+export declare const reportCases: readonly [{
+    readonly name: "cost split and unresolved evidence are surfaced";
+    readonly peerCost: 14.652426;
+    readonly generationCost: 1.876718;
+    readonly totalCost: 16.529144;
+    readonly unresolvedAsk: "attach raw npm test output";
+}];

package/dist/scripts/eval-fixtures.js

@@ -0,0 +1,216 @@
+import assert from "node:assert/strict";
+import fs from "node:fs";
+import os from "node:os";
+import path from "node:path";
+import { loadConfig } from "../src/core/config.js";
+import { truthfulnessPreflight } from "../src/core/orchestrator.js";
+import { sessionReportMarkdown } from "../src/core/reports.js";
+import { SessionStore } from "../src/core/session-store.js";
+import { parsePeerStatus } from "../src/core/status.js";
+function evalTmpDir(label) {
+    return fs.mkdtempSync(path.join(os.tmpdir(), `cross-review-eval-${label}-`));
+}
+export const truthfulnessCases = [
+    {
+        name: "current runtime contradiction is blocked",
+        input: {
+            task: "The current cross-review runtime is 4.2.4.",
+            runtimeFacts: { runtime_version: "4.2.5", release_date: "2026-06-05" },
+            attachmentsPresent: false,
+        },
+        expectPass: false,
+        expectIssueClass: "runtime_contradiction",
+    },
+    {
+        name: "matching current runtime facts pass",
+        input: {
+            task: "server_info shows current cross-review runtime 4.2.5.",
+            runtimeFacts: { runtime_version: "4.2.5", release_date: "2026-06-05" },
+            attachmentsPresent: false,
+        },
+        expectPass: true,
+    },
+    {
+        name: "historical timing claim needs snapshot evidence",
+        input: {
+            task: "When the audit began, cross-review was running 4.2.4.",
+            runtimeFacts: { runtime_version: "4.2.5", release_date: "2026-06-05" },
+            attachmentsPresent: false,
+        },
+        expectPass: false,
+        expectIssueClass: "unsupported_historical_claim",
+    },
+];
+export const parserCases = [
+    {
+        name: "verified with empty evidence gets empty-evidence warning",
+        text: JSON.stringify({
+            status: "READY",
+            summary: "ok",
+            confidence: "verified",
+            evidence_sources: [],
+            caller_requests: [],
+            follow_ups: [],
+        }),
+        expectStatus: "READY",
+        expectWarning: "verified_without_evidence_sources",
+    },
+    {
+        name: "verified with attached evidence path is concrete",
+        text: JSON.stringify({
+            status: "READY",
+            summary: "ok",
+            confidence: "verified",
+            evidence_sources: ["evidence/2026-06-05T00-00-00Z-raw-smoke.txt: npm test 42 passed"],
+            caller_requests: [],
+            follow_ups: [],
+        }),
+        expectStatus: "READY",
+        absentWarning: "verified_without_evidence_sources",
+    },
+];
+export const reportCases = [
+    {
+        name: "cost split and unresolved evidence are surfaced",
+        peerCost: 14.652426,
+        generationCost: 1.876718,
+        totalCost: 16.529144,
+        unresolvedAsk: "attach raw npm test output",
+    },
+];
+for (const testCase of truthfulnessCases) {
+    const result = truthfulnessPreflight({
+        task: testCase.input.task,
+        runtimeFacts: testCase.input.runtimeFacts,
+        attachmentsPresent: testCase.input.attachmentsPresent,
+    });
+    assert.equal(result.pass, testCase.expectPass, testCase.name);
+    if ("expectIssueClass" in testCase) {
+        assert.ok(result.issue_classes.includes(testCase.expectIssueClass), testCase.name);
+    }
+}
+for (const testCase of parserCases) {
+    const result = parsePeerStatus(testCase.text);
+    assert.equal(result.status, testCase.expectStatus, testCase.name);
+    if ("expectWarning" in testCase) {
+        assert.ok(result.parser_warnings.includes(testCase.expectWarning), testCase.name);
+    }
+    if ("absentWarning" in testCase) {
+        assert.ok(!result.parser_warnings.includes(testCase.absentWarning), testCase.name);
+    }
+}
+for (const testCase of reportCases) {
+    const store = new SessionStore({
+        ...loadConfig(),
+        data_dir: evalTmpDir("report"),
+    });
+    const session = await store.init(`eval report fixture: ${testCase.name}`, "operator", []);
+    const meta = store.read(session.session_id);
+    const ts = new Date().toISOString();
+    meta.rounds = [
+        {
+            round: 1,
+            started_at: ts,
+            completed_at: ts,
+            caller_status: "READY",
+            prompt_file: "agent-runs/round-1-prompt.md",
+            peers: [
+                {
+                    peer: "codex",
+                    provider: "openai",
+                    model: "gpt-5.5",
+                    status: "READY",
+                    structured: {
+                        status: "READY",
+                        summary: "ready",
+                        confidence: "verified",
+                        evidence_sources: ["server_info: version 4.2.5"],
+                        caller_requests: [],
+                        follow_ups: [],
+                    },
+                    text: "{}",
+                    raw: { fixture: true },
+                    decision_quality: "clean",
+                    parser_warnings: [],
+                    attempts: 1,
+                    latency_ms: 1,
+                    usage: { input_tokens: 1, output_tokens: 1, total_tokens: 2 },
+                    cost: {
+                        currency: "USD",
+                        estimated: false,
+                        source: "configured-rate",
+                        total_cost: testCase.peerCost,
+                    },
+                },
+            ],
+            rejected: [],
+            convergence: {
+                converged: true,
+                reason: "fixture",
+                ready_peers: ["codex"],
+                not_ready_peers: [],
+                needs_evidence_peers: [],
+                rejected_peers: [],
+                skipped_peers: [],
+                decision_quality: {
+                    codex: "clean",
+                    claude: "clean",
+                    gemini: "clean",
+                    deepseek: "clean",
+                    grok: "clean",
+                    perplexity: "clean",
+                },
+                blocking_details: [],
+            },
+        },
+    ];
+    meta.generation_files = [
+        {
+            round: 0,
+            peer: "codex",
+            label: "initial_draft",
+            path: "agent-runs/round-0-initial-draft.md",
+            ts,
+            usage: { input_tokens: 1, output_tokens: 1, total_tokens: 2 },
+            cost: {
+                currency: "USD",
+                estimated: false,
+                source: "configured-rate",
+                total_cost: testCase.generationCost,
+            },
+        },
+    ];
+    meta.totals.cost = {
+        currency: "USD",
+        estimated: false,
+        source: "configured-rate",
+        total_cost: testCase.totalCost,
+    };
+    meta.evidence_checklist = [
+        {
+            id: "eval-1",
+            peer: "codex",
+            first_round: 1,
+            last_round: 1,
+            round_count: 1,
+            ask: testCase.unresolvedAsk,
+            first_seen_at: ts,
+            last_seen_at: ts,
+            status: "not_resurfaced",
+            addressed_at_round: 2,
+            address_method: "resurfacing",
+        },
+    ];
+    fs.writeFileSync(store.metaPath(session.session_id), JSON.stringify(meta));
+    const report = sessionReportMarkdown(store.read(session.session_id), []);
+    assert.ok(report.includes("$16.529144 USD = $14.652426 peer + $1.876718 generation"));
+    assert.ok(report.includes("## Unresolved Evidence Disposition"));
+    assert.ok(report.includes(testCase.unresolvedAsk));
+}
+console.log(JSON.stringify({
+    ok: true,
+    truthfulness_cases: truthfulnessCases.length,
+    parser_cases: parserCases.length,
+    report_cases: reportCases.length,
+}));
+//# sourceMappingURL=eval-fixtures.js.map

package/dist/scripts/eval-fixtures.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"eval-fixtures.js","sourceRoot":"","sources":["../../scripts/eval-fixtures.ts"],"names":[],"mappings":"AAAA,OAAO,MAAM,MAAM,oBAAoB,CAAC;AACxC,OAAO,EAAE,MAAM,SAAS,CAAC;AACzB,OAAO,EAAE,MAAM,SAAS,CAAC;AACzB,OAAO,IAAI,MAAM,WAAW,CAAC;AAC7B,OAAO,EAAE,UAAU,EAAE,MAAM,uBAAuB,CAAC;AACnD,OAAO,EAAE,qBAAqB,EAAE,MAAM,6BAA6B,CAAC;AACpE,OAAO,EAAE,qBAAqB,EAAE,MAAM,wBAAwB,CAAC;AAC/D,OAAO,EAAE,YAAY,EAAE,MAAM,8BAA8B,CAAC;AAC5D,OAAO,EAAE,eAAe,EAAE,MAAM,uBAAuB,CAAC;AAExD,SAAS,UAAU,CAAC,KAAa;IAC/B,OAAO,EAAE,CAAC,WAAW,CAAC,IAAI,CAAC,IAAI,CAAC,EAAE,CAAC,MAAM,EAAE,EAAE,qBAAqB,KAAK,GAAG,CAAC,CAAC,CAAC;AAC/E,CAAC;AAED,MAAM,CAAC,MAAM,iBAAiB,GAAG;IAC/B;QACE,IAAI,EAAE,0CAA0C;QAChD,KAAK,EAAE;YACL,IAAI,EAAE,4CAA4C;YAClD,YAAY,EAAE,EAAE,eAAe,EAAE,OAAO,EAAE,YAAY,EAAE,YAAY,EAAE;YACtE,kBAAkB,EAAE,KAAK;SAC1B;QACD,UAAU,EAAE,KAAK;QACjB,gBAAgB,EAAE,uBAAuB;KAC1C;IACD;QACE,IAAI,EAAE,qCAAqC;QAC3C,KAAK,EAAE;YACL,IAAI,EAAE,uDAAuD;YAC7D,YAAY,EAAE,EAAE,eAAe,EAAE,OAAO,EAAE,YAAY,EAAE,YAAY,EAAE;YACtE,kBAAkB,EAAE,KAAK;SAC1B;QACD,UAAU,EAAE,IAAI;KACjB;IACD;QACE,IAAI,EAAE,iDAAiD;QACvD,KAAK,EAAE;YACL,IAAI,EAAE,uDAAuD;YAC7D,YAAY,EAAE,EAAE,eAAe,EAAE,OAAO,EAAE,YAAY,EAAE,YAAY,EAAE;YACtE,kBAAkB,EAAE,KAAK;SAC1B;QACD,UAAU,EAAE,KAAK;QACjB,gBAAgB,EAAE,8BAA8B;KACjD;CACO,CAAC;AAEX,MAAM,CAAC,MAAM,WAAW,GAAG;IACzB;QACE,IAAI,EAAE,0DAA0D;QAChE,IAAI,EAAE,IAAI,CAAC,SAAS,CAAC;YACnB,MAAM,EAAE,OAAO;YACf,OAAO,EAAE,IAAI;YACb,UAAU,EAAE,UAAU;YACtB,gBAAgB,EAAE,EAAE;YACpB,eAAe,EAAE,EAAE;YACnB,UAAU,EAAE,EAAE;SACf,CAAC;QACF,YAAY,EAAE,OAAO;QACrB,aAAa,EAAE,mCAAmC;KACnD;IACD;QACE,IAAI,EAAE,kDAAkD;QACxD,IAAI,EAAE,IAAI,CAAC,SAAS,CAAC;YACnB,MAAM,EAAE,OAAO;YACf,OAAO,EAAE,IAAI;YACb,UAAU,EAAE,UAAU;YACtB,gBAAgB,EAAE,CAAC,iEAAiE,CAAC;YACrF,eAAe,EAAE,EAAE;YACnB,UAAU,EAAE,EAAE;SACf,CAAC;QACF,YAAY,EAAE,OAAO;QACrB,aAAa,EAAE,mCAAmC;KACnD;CACO,CAAC;AAEX,MAAM,CAAC,MAAM,WAAW,GAAG;IACzB;QACE,IAAI,EAAE,iDAAiD;QACvD,QAAQ,EAAE,SAAS;QACnB,cAAc,EAAE,QAAQ;QACxB,SAAS,EAAE,SAAS;QACpB,aAAa,EAAE,4BAA4B;KAC5C;CACO,CAAC;AAEX,KAAK,MAAM,QAAQ,IAAI,iBAAiB,EAAE,CAAC;IACzC,MAAM,MAAM,GAAG,qBAAqB,CAAC;QACnC,IAAI,EAAE,QAAQ,CAAC,KAAK,CAAC,IAAI;QACzB,YAAY,EAAE,QAAQ,CAAC,KAAK,CAAC,YAAY;QACzC,kBAAkB,EAAE,QAAQ,CAAC,KAAK,CAAC,kBAAkB;KACtD,CAAC,CAAC;IACH,MAAM,CAAC,KAAK,CAAC,MAAM,CAAC,IAAI,EAAE,QAAQ,CAAC,UAAU,EAAE,QAAQ,CAAC,IAAI,CAAC,CAAC;IAC9D,IAAI,kBAAkB,IAAI,QAAQ,EAAE,CAAC;QACnC,MAAM,CAAC,EAAE,CAAC,MAAM,CAAC,aAAa,CAAC,QAAQ,CAAC,QAAQ,CAAC,gBAAgB,CAAC,EAAE,QAAQ,CAAC,IAAI,CAAC,CAAC;IACrF,CAAC;AACH,CAAC;AAED,KAAK,MAAM,QAAQ,IAAI,WAAW,EAAE,CAAC;IACnC,MAAM,MAAM,GAAG,eAAe,CAAC,QAAQ,CAAC,IAAI,CAAC,CAAC;IAC9C,MAAM,CAAC,KAAK,CAAC,MAAM,CAAC,MAAM,EAAE,QAAQ,CAAC,YAAY,EAAE,QAAQ,CAAC,IAAI,CAAC,CAAC;IAClE,IAAI,eAAe,IAAI,QAAQ,EAAE,CAAC;QAChC,MAAM,CAAC,EAAE,CAAC,MAAM,CAAC,eAAe,CAAC,QAAQ,CAAC,QAAQ,CAAC,aAAa,CAAC,EAAE,QAAQ,CAAC,IAAI,CAAC,CAAC;IACpF,CAAC;IACD,IAAI,eAAe,IAAI,QAAQ,EAAE,CAAC;QAChC,MAAM,CAAC,EAAE,CAAC,CAAC,MAAM,CAAC,eAAe,CAAC,QAAQ,CAAC,QAAQ,CAAC,aAAa,CAAC,EAAE,QAAQ,CAAC,IAAI,CAAC,CAAC;IACrF,CAAC;AACH,CAAC;AAED,KAAK,MAAM,QAAQ,IAAI,WAAW,EAAE,CAAC;IACnC,MAAM,KAAK,GAAG,IAAI,YAAY,CAAC;QAC7B,GAAG,UAAU,EAAE;QACf,QAAQ,EAAE,UAAU,CAAC,QAAQ,CAAC;KAC/B,CAAC,CAAC;IACH,MAAM,OAAO,GAAG,MAAM,KAAK,CAAC,IAAI,CAAC,wBAAwB,QAAQ,CAAC,IAAI,EAAE,EAAE,UAAU,EAAE,EAAE,CAAC,CAAC;IAC1F,MAAM,IAAI,GAAG,KAAK,CAAC,IAAI,CAAC,OAAO,CAAC,UAAU,CAAC,CAAC;IAC5C,MAAM,EAAE,GAAG,IAAI,IAAI,EAAE,CAAC,WAAW,EAAE,CAAC;IACpC,IAAI,CAAC,MAAM,GAAG;QACZ;YACE,KAAK,EAAE,CAAC;YACR,UAAU,EAAE,EAAE;YACd,YAAY,EAAE,EAAE;YAChB,aAAa,EAAE,OAAO;YACtB,WAAW,EAAE,8BAA8B;YAC3C,KAAK,EAAE;gBACL;oBACE,IAAI,EAAE,OAAO;oBACb,QAAQ,EAAE,QAAQ;oBAClB,KAAK,EAAE,SAAS;oBAChB,MAAM,EAAE,OAAO;oBACf,UAAU,EAAE;wBACV,MAAM,EAAE,OAAO;wBACf,OAAO,EAAE,OAAO;wBAChB,UAAU,EAAE,UAAU;wBACtB,gBAAgB,EAAE,CAAC,4BAA4B,CAAC;wBAChD,eAAe,EAAE,EAAE;wBACnB,UAAU,EAAE,EAAE;qBACf;oBACD,IAAI,EAAE,IAAI;oBACV,GAAG,EAAE,EAAE,OAAO,EAAE,IAAI,EAAE;oBACtB,gBAAgB,EAAE,OAAO;oBACzB,eAAe,EAAE,EAAE;oBACnB,QAAQ,EAAE,CAAC;oBACX,UAAU,EAAE,CAAC;oBACb,KAAK,EAAE,EAAE,YAAY,EAAE,CAAC,EAAE,aAAa,EAAE,CAAC,EAAE,YAAY,EAAE,CAAC,EAAE;oBAC7D,IAAI,EAAE;wBACJ,QAAQ,EAAE,KAAK;wBACf,SAAS,EAAE,KAAK;wBAChB,MAAM,EAAE,iBAAiB;wBACzB,UAAU,EAAE,QAAQ,CAAC,QAAQ;qBAC9B;iBACF;aACF;YACD,QAAQ,EAAE,EAAE;YACZ,WAAW,EAAE;gBACX,SAAS,EAAE,IAAI;gBACf,MAAM,EAAE,SAAS;gBACjB,WAAW,EAAE,CAAC,OAAO,CAAC;gBACtB,eAAe,EAAE,EAAE;gBACnB,oBAAoB,EAAE,EAAE;gBACxB,cAAc,EAAE,EAAE;gBAClB,aAAa,EAAE,EAAE;gBACjB,gBAAgB,EAAE;oBAChB,KAAK,EAAE,OAAO;oBACd,MAAM,EAAE,OAAO;oBACf,MAAM,EAAE,OAAO;oBACf,QAAQ,EAAE,OAAO;oBACjB,IAAI,EAAE,OAAO;oBACb,UAAU,EAAE,OAAO;iBACpB;gBACD,gBAAgB,EAAE,EAAE;aACrB;SACF;KACF,CAAC;IACF,IAAI,CAAC,gBAAgB,GAAG;QACtB;YACE,KAAK,EAAE,CAAC;YACR,IAAI,EAAE,OAAO;YACb,KAAK,EAAE,eAAe;YACtB,IAAI,EAAE,qCAAqC;YAC3C,EAAE;YACF,KAAK,EAAE,EAAE,YAAY,EAAE,CAAC,EAAE,aAAa,EAAE,CAAC,EAAE,YAAY,EAAE,CAAC,EAAE;YAC7D,IAAI,EAAE;gBACJ,QAAQ,EAAE,KAAK;gBACf,SAAS,EAAE,KAAK;gBAChB,MAAM,EAAE,iBAAiB;gBACzB,UAAU,EAAE,QAAQ,CAAC,cAAc;aACpC;SACF;KACF,CAAC;IACF,IAAI,CAAC,MAAM,CAAC,IAAI,GAAG;QACjB,QAAQ,EAAE,KAAK;QACf,SAAS,EAAE,KAAK;QAChB,MAAM,EAAE,iBAAiB;QACzB,UAAU,EAAE,QAAQ,CAAC,SAAS;KAC/B,CAAC;IACF,IAAI,CAAC,kBAAkB,GAAG;QACxB;YACE,EAAE,EAAE,QAAQ;YACZ,IAAI,EAAE,OAAO;YACb,WAAW,EAAE,CAAC;YACd,UAAU,EAAE,CAAC;YACb,WAAW,EAAE,CAAC;YACd,GAAG,EAAE,QAAQ,CAAC,aAAa;YAC3B,aAAa,EAAE,EAAE;YACjB,YAAY,EAAE,EAAE;YAChB,MAAM,EAAE,gBAAgB;YACxB,kBAAkB,EAAE,CAAC;YACrB,cAAc,EAAE,aAAa;SAC9B;KACF,CAAC;IACF,EAAE,CAAC,aAAa,CAAC,KAAK,CAAC,QAAQ,CAAC,OAAO,CAAC,UAAU,CAAC,EAAE,IAAI,CAAC,SAAS,CAAC,IAAI,CAAC,CAAC,CAAC;IAC3E,MAAM,MAAM,GAAG,qBAAqB,CAAC,KAAK,CAAC,IAAI,CAAC,OAAO,CAAC,UAAU,CAAC,EAAE,EAAE,CAAC,CAAC;IACzE,MAAM,CAAC,EAAE,CAAC,MAAM,CAAC,QAAQ,CAAC,yDAAyD,CAAC,CAAC,CAAC;IACtF,MAAM,CAAC,EAAE,CAAC,MAAM,CAAC,QAAQ,CAAC,oCAAoC,CAAC,CAAC,CAAC;IACjE,MAAM,CAAC,EAAE,CAAC,MAAM,CAAC,QAAQ,CAAC,QAAQ,CAAC,aAAa,CAAC,CAAC,CAAC;AACrD,CAAC;AAED,OAAO,CAAC,GAAG,CACT,IAAI,CAAC,SAAS,CAAC;IACb,EAAE,EAAE,IAAI;IACR,kBAAkB,EAAE,iBAAiB,CAAC,MAAM;IAC5C,YAAY,EAAE,WAAW,CAAC,MAAM;IAChC,YAAY,EAAE,WAAW,CAAC,MAAM;CACjC,CAAC,CACH,CAAC"}

package/dist/scripts/smoke.js

@@ -22,6 +22,7 @@ import { SWEEP_MIN_IDLE_MS } from "../src/core/session-store.js";
 import { parsePeerStatus } from "../src/core/status.js";
 import { PEERS } from "../src/core/types.js";
 import { getCallerCandidatesFromClientInfo, lockCallerPeerSelection, pruneCompletedJobs, SessionIdSchema, setHostTokensRecord, verifyCallerIdentity, } from "../src/mcp/server.js";
+import { classifyProviderError } from "../src/peers/errors.js";
 import { selectFromCandidates } from "../src/peers/model-selection.js";
 import { StubAdapter } from "../src/peers/stub.js";
 import { redact } from "../src/security/redact.js";
@@ -488,13 +489,13 @@ assert.equal(checkConvergence(["codex", "claude"], "READY", [fakeReady("codex"),
 // taxonomy + the skip-gated quorum floor + the zero-skip non-regression
 // invariant.
 {
-    const fakeFailure = (peer, failureClass) => ({
+    const fakeFailure = (peer, failureClass, retryable = true) => ({
         peer,
         provider: "stub",
         model: "stub",
         failure_class: failureClass,
         message: `stub ${failureClass}`,
-        retryable: true,
+        retryable,
         attempts: 3,
         latency_ms: 0,
     });
@@ -509,8 +510,12 @@ assert.equal(checkConvergence(["codex", "claude"], "READY", [fakeReady("codex"),
         "fallback_exhausted",
     ]) {
         assert.equal(isSkippableFailure(fakeFailure("grok", fc)), true, `v3.7.3 / skip-peer: ${fc} must be skippable (infra unavailability)`);
-        assert.equal(SKIPPABLE_FAILURE_CLASSES.has(fc), true, `v3.7.3 / skip-peer: ${fc} must be in SKIPPABLE_FAILURE_CLASSES`);
     }
+    assert.equal(SKIPPABLE_FAILURE_CLASSES.has("provider_error"), true, "v3.7.3 / skip-peer: provider_error remains a known skip candidate when retryable");
+    assert.equal(isSkippableFailure(fakeFailure("grok", "provider_error", false)), false, "v4.3.1 / skip-peer: non-retryable provider_error (e.g. provider 400 payload/schema rejection) must block instead of being skipped");
+    const anthropicOverloaded = classifyProviderError("claude", "anthropic", "claude-opus-4-8", new Error('{"type":"error","error":{"details":null,"type":"overloaded_error","message":"Overloaded"},"request_id":"req_fixture"}'), 1, Date.now());
+    assert.equal(anthropicOverloaded.failure_class, "provider_error");
+    assert.equal(anthropicOverloaded.retryable, true, "v4.3.1 / provider-errors: Anthropic overloaded_error without HTTP status text must still be retryable");
     for (const fc of [
         "schema",
         "unparseable_after_recovery",
@@ -1248,6 +1253,193 @@ assert.equal(Object.hasOwn(metrics.decision_quality, "undefined"), false);
     assert.ok(notResurfacedReport.includes("not_resurfaced means the ask was not repeated; it is not proof that evidence was satisfied."), "v4.2.5 / not_resurfaced: session report must state the not_resurfaced semantics");
     console.log("[smoke] terminal_cost_evidence_audit_test: PASS");
 }
+// v4.3.0 / P1: unanimous READY with unresolved evidence must not look like a
+// plain unanimous_ready close-out. `not_resurfaced` is inference-only: it may
+// allow convergence, but the final metadata/report must keep that disposition
+// visible for operators.
+{
+    const { sessionReportMarkdown } = await import("../src/core/reports.js");
+    const unresolvedEvents = [];
+    const unresolvedConfig = {
+        ...loadConfig(),
+        data_dir: smokeTmpDir("unresolved-evidence-finalize"),
+        budget: {
+            ...loadConfig().budget,
+            max_session_cost_usd: 10000,
+            preflight_max_round_cost_usd: 10000,
+            until_stopped_max_cost_usd: 10000,
+        },
+        evidence_judge_autowire: {
+            ...loadConfig().evidence_judge_autowire,
+            mode: "off",
+            active: false,
+        },
+    };
+    const unresolvedOrch = new CrossReviewOrchestrator(unresolvedConfig, (event) => unresolvedEvents.push(event.type));
+    const unresolvedR1 = await unresolvedOrch.askPeers({
+        task: "P1 unresolved evidence finalization guard fixture.",
+        draft: "FORCE_NEEDS_EVIDENCE",
+        caller: "operator",
+        peers: ["claude"],
+    });
+    const unresolvedR2 = await unresolvedOrch.askPeers({
+        session_id: unresolvedR1.session.session_id,
+        task: "P1 unresolved evidence finalization guard fixture.",
+        draft: "Clean revised draft, no test marker present.",
+        caller: "operator",
+        peers: ["claude"],
+    });
+    assert.equal(unresolvedR2.converged, true);
+    assert.equal(unresolvedR2.session.outcome, "converged");
+    assert.equal(unresolvedR2.session.outcome_reason, "unanimous_ready_with_unresolved_evidence", "v4.3.0 / P1: convergence with not_resurfaced evidence must not finalize as plain unanimous_ready");
+    assert.ok(unresolvedEvents.includes("session.evidence_checklist_unresolved_on_finalize"), "v4.3.0 / P1: unresolved evidence close-out must emit an audit event");
+    const unresolvedReport = sessionReportMarkdown(unresolvedOrch.store.read(unresolvedR2.session.session_id), unresolvedOrch.store.readEvents(unresolvedR2.session.session_id));
+    assert.ok(unresolvedReport.includes("## Unresolved Evidence Disposition"), "v4.3.0 / P1: session_report must include unresolved-evidence disposition table");
+    assert.ok(unresolvedReport.includes("not_resurfaced"), "v4.3.0 / P1: session_report must name not_resurfaced unresolved items");
+    console.log("[smoke] unresolved_evidence_finalization_guard_test: PASS");
+}
+// v4.3.0 / P3: read-only peer reliability telemetry. This is deliberately
+// observational; it must not change peer selection or mutate sessions.
+{
+    const { SessionStore } = await import("../src/core/session-store.js");
+    const reliabilityStore = new SessionStore({
+        ...config,
+        data_dir: smokeTmpDir("peer-reliability"),
+    });
+    const reliabilitySession = await reliabilityStore.init("peer reliability report fixture", "operator", []);
+    const reliabilityMeta = reliabilityStore.read(reliabilitySession.session_id);
+    const ts = new Date().toISOString();
+    reliabilityMeta.rounds = [
+        {
+            round: 1,
+            started_at: ts,
+            completed_at: ts,
+            caller_status: "READY",
+            prompt_file: "agent-runs/round-1-prompt.md",
+            peers: [
+                {
+                    peer: "claude",
+                    provider: "anthropic",
+                    model: "claude-opus-4-8",
+                    status: "NEEDS_EVIDENCE",
+                    structured: {
+                        status: "NEEDS_EVIDENCE",
+                        summary: "needs log",
+                        confidence: "verified",
+                        evidence_sources: ["src/core/session-store.ts:1"],
+                        caller_requests: ["attach raw npm test output"],
+                        follow_ups: [],
+                    },
+                    text: "{}",
+                    raw: { fixture: true },
+                    decision_quality: "clean",
+                    parser_warnings: [],
+                    attempts: 1,
+                    latency_ms: 50,
+                    usage: { input_tokens: 10, output_tokens: 5, total_tokens: 15 },
+                    cost: { currency: "USD", estimated: false, source: "configured-rate", total_cost: 1 },
+                },
+                {
+                    peer: "grok",
+                    provider: "xai",
+                    model: "grok-4.3",
+                    status: "READY",
+                    structured: {
+                        status: "READY",
+                        summary: "ready",
+                        confidence: "verified",
+                        evidence_sources: ["server_info: version 4.2.5"],
+                        caller_requests: [],
+                        follow_ups: [],
+                    },
+                    text: "{}",
+                    raw: { fixture: true },
+                    decision_quality: "format_warning",
+                    parser_warnings: ["verified_without_concrete_evidence_sources"],
+                    attempts: 1,
+                    latency_ms: 100,
+                    usage: { input_tokens: 20, output_tokens: 10, total_tokens: 30 },
+                    cost: { currency: "USD", estimated: false, source: "configured-rate", total_cost: 2 },
+                },
+            ],
+            rejected: [
+                {
+                    peer: "perplexity",
+                    provider: "perplexity",
+                    model: "sonar-reasoning-pro",
+                    failure_class: "provider_error",
+                    message: "fixture provider error",
+                    retryable: false,
+                    attempts: 1,
+                    latency_ms: 0,
+                },
+            ],
+            convergence: {
+                converged: false,
+                reason: "fixture",
+                ready_peers: ["grok"],
+                not_ready_peers: [],
+                needs_evidence_peers: ["claude"],
+                rejected_peers: ["perplexity"],
+                skipped_peers: [],
+                decision_quality: {
+                    codex: "clean",
+                    claude: "clean",
+                    gemini: "clean",
+                    deepseek: "clean",
+                    grok: "format_warning",
+                    perplexity: "failed",
+                },
+                blocking_details: ["claude:NEEDS_EVIDENCE", "perplexity:provider_error"],
+            },
+        },
+    ];
+    reliabilityMeta.evidence_checklist = [
+        {
+            id: "rel-1",
+            peer: "claude",
+            first_round: 1,
+            last_round: 1,
+            round_count: 1,
+            ask: "attach raw npm test output",
+            first_seen_at: ts,
+            last_seen_at: ts,
+            status: "not_resurfaced",
+            addressed_at_round: 2,
+            address_method: "resurfacing",
+        },
+    ];
+    fs.writeFileSync(reliabilityStore.metaPath(reliabilitySession.session_id), JSON.stringify(reliabilityMeta));
+    await reliabilityStore.appendEvent({
+        ts,
+        type: "session.lead_meta_audit_fabrication_detected",
+        session_id: reliabilitySession.session_id,
+        message: "fixture fabrication event",
+        data: { peer: "grok" },
+    });
+    const reliability = reliabilityStore.peerReliabilityReport();
+    assert.equal(reliability.scope, "all");
+    assert.equal(reliability.by_peer.claude?.needs_evidence, 1);
+    assert.equal(reliability.by_peer.claude?.not_resurfaced_asks, 1);
+    assert.equal(reliability.by_peer.grok?.ready, 1);
+    assert.equal(reliability.by_peer.grok?.parser_warnings_total, 1);
+    assert.equal(reliability.by_peer.grok?.fabrication_events, 1);
+    assert.equal(reliability.by_peer.perplexity?.provider_errors, 1);
+    console.log("[smoke] peer_reliability_report_test: PASS");
+}
+// v4.3.0 / P2: offline declarative eval harness. This pins the existence of a
+// no-provider-call fixture runner so regressions found in real sessions can be
+// replayed without growing the ad hoc smoke body indefinitely.
+{
+    const pkg = JSON.parse(fs.readFileSync("package.json", "utf8"));
+    assert.equal(pkg.scripts?.["eval:fixtures"], "tsx scripts/eval-fixtures.ts", "v4.3.0 / P2: package.json must expose the offline fixture eval runner");
+    const evalHarness = fs.readFileSync("scripts/eval-fixtures.ts", "utf8");
+    assert.ok(/truthfulnessCases/.test(evalHarness) &&
+        /parserCases/.test(evalHarness) &&
+        /reportCases/.test(evalHarness), "v4.3.0 / P2: eval-fixtures must use declarative truthfulness/parser/report case tables");
+    assert.ok(!/askPeers\(|runUntilUnanimous\(|session_start_round/.test(evalHarness), "v4.3.0 / P2: eval-fixtures must stay offline and avoid provider-review entry points");
+    console.log("[smoke] offline_fixture_eval_contract_test: PASS");
+}
 // v2.22.0 (B.P3): session.budget_warning event emit + idempotency. The
 // orchestrator emits a one-shot warning when cumulative cost crosses
 // 75% of cost_ceiling_usd; the budget_warning_emitted flag persists