@rulvar/evals 1.189.0 → 1.191.0

package/dist/index.d.ts

@@ -838,8 +838,14 @@ declare const FAULT_SCENARIO_NAMES: readonly string[];
 declare function runFaultInjection(options?: RunFaultInjectionOptions): Promise<FaultInjectionReport>;
 //#endregion
 //#region src/claim-corpus.d.ts
-/** The failure classes the eighteenth benchmark shipped, plus the bound classes. */
-type ClaimCorpusClass = "live-fact" | "package-identity" | "inverted-default" | "numeric-range" | "negation" | "bounded-coverage";
+/**
+* The failure classes the eighteenth benchmark shipped, plus the bound
+* classes, plus the nineteenth benchmark's pair (RV1809):
+* 'modality-overclaim' is a mitigation stated as an unconditional
+* guarantee, and 'scope-ambiguity' is a child-only total printed as a
+* whole-workflow figure.
+*/
+type ClaimCorpusClass = "live-fact" | "package-identity" | "inverted-default" | "numeric-range" | "negation" | "bounded-coverage" | "modality-overclaim" | "scope-ambiguity";
 /** One adversarial case: a draft, its contradicting evidence, and the mechanical expectations. */
 interface ClaimCorpusCase {
   id: string;

package/dist/index.js

@@ -2748,6 +2748,40 @@ const CLAIM_CORPUS = [
 			minPairs: 1,
 			coverage: "partial"
 		}
+	},
+	{
+		id: "modality-overclaim-attestation-stop",
+		class: "modality-overclaim",
+		draft: "Any drift of an attested toolset, the executable body included, is stopped pre-wire unconditionally: a changed tool always refuses at spawn time (packages/core/src/tools/toolset-hash.ts:85).",
+		pool: [{
+			nodeId: "agent:3",
+			text: "The attestation pin holds the resolved toolset to the attested CONTRACT hash, name, description, parameters, and version (packages/core/src/tools/toolset-hash.ts:85). An executable body edit under an unchanged contract does not move that hash by design; version is the drift lever, and the authority hash covers risk, needsApproval, executor, and executorSpec (packages/core/src/l0/spi/toolsource.ts:53)."
+		}],
+		critical: ["packages/core/src/tools/toolset-hash.ts"],
+		expect: {
+			minPairs: 1,
+			anchors: ["packages/core/src/tools/toolset-hash.ts:85"],
+			coverage: "full"
+		}
+	},
+	{
+		id: "scope-ambiguity-child-totals-as-workflow",
+		class: "scope-ambiguity",
+		draft: "The current workflow performed exactly 100 wire requests in total, 6126893 input tokens and 94555 output tokens end to end.",
+		runFacts: {
+			text: "RUN FACTS, scope settled-children-only: 100 wire requests, 6126893 input tokens and 94555 output tokens across the six settled children; the terminal workflow invoice additionally carries the orchestrator, the judges, and the synthesis at 118 wires, 6517187 input and 138947 output.",
+			ids: ["comparison-run-aug04"],
+			numbers: [
+				100,
+				6126893,
+				94555,
+				118,
+				6517187,
+				138947
+			]
+		},
+		runFactTerms: ["in total", "end to end"],
+		expect: { minRunFactPairs: 1 }
 	}
 ];
 /**

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@rulvar/evals",
-  "version": "1.189.0",
+  "version": "1.191.0",
   "description": "Rulvar evals: eval cases, golden outputs, rubric and judge graders, matrix sweeps, canary fingerprint.",
   "type": "module",
   "license": "Apache-2.0",
@@ -22,11 +22,11 @@
     "access": "public"
   },
   "dependencies": {
-    "@rulvar/anthropic": "1.189.0",
-    "@rulvar/openai": "1.189.0",
-    "@rulvar/plan": "1.189.0",
-    "@rulvar/core": "1.189.0",
-    "@rulvar/testing": "1.189.0"
+    "@rulvar/core": "1.191.0",
+    "@rulvar/anthropic": "1.191.0",
+    "@rulvar/openai": "1.191.0",
+    "@rulvar/plan": "1.191.0",
+    "@rulvar/testing": "1.191.0"
   },
   "devDependencies": {
     "@types/node": "^22.20.1",