@intentsolutions/jrig-cli 0.1.0 → 0.1.1

package/dist/index.js

@@ -82,6 +82,9 @@ var CriterionSchema = z.object({
   deterministic_check_params: z.record(z.string(), z.unknown()).optional().describe(
     "Parameters forwarded to the deterministic check (e.g. { value: 'needle' } for 'contains', { pattern: '\\\\d+' } for 'regex_match')"
   )
+}).refine((c) => c.method !== "deterministic" || !!c.deterministic_check, {
+  message: "deterministic criteria must define deterministic_check",
+  path: ["deterministic_check"]
 });
 var TestCaseTier = z2.enum(["core", "edge", "regression", "adversarial"]);
 var TriggerExpectation = z2.enum(["should_trigger", "should_not_trigger"]);
@@ -115,6 +118,19 @@ var EvalSpecSchema = z3.object({
   models: z3.array(ModelTarget).default(["sonnet"]).describe("Models to test independently"),
   siblings: z3.array(SiblingSkillSchema).optional().describe("Sibling skills for pack-sensitive evaluation"),
   tags: z3.array(z3.string()).optional().describe("Categorization tags")
+}).superRefine((spec, ctx) => {
+  const knownCriteria = new Set(spec.criteria.map((c) => c.id));
+  spec.test_cases.forEach((tc, ti) => {
+    tc.criteria_ids?.forEach((cid, ci) => {
+      if (!knownCriteria.has(cid)) {
+        ctx.addIssue({
+          code: z3.ZodIssueCode.custom,
+          message: `test case "${tc.id}" references unknown criterion id "${cid}"`,
+          path: ["test_cases", ti, "criteria_ids", ci]
+        });
+      }
+    });
+  });
 });
 var EvalContractSchema = z4.object({
   contract_version: z4.literal("1.0").describe("Schema version for forward compatibility"),
@@ -1009,6 +1025,16 @@ async function judgeWithLLM(criterion, outcome, provider, model) {
     };
   }
 }
+function selectCriteriaForTestCase(criteria, criteriaIds) {
+  if (criteriaIds === void 0) return criteria;
+  const available = new Set(criteria.map((c) => c.id));
+  const unknown = criteriaIds.filter((id) => !available.has(id));
+  if (unknown.length > 0) {
+    throw new Error(`Test case references unknown criteria_ids: ${unknown.join(", ")}`);
+  }
+  const wanted = new Set(criteriaIds);
+  return criteria.filter((c) => wanted.has(c.id));
+}
 function computeScoreCard(results, criteria, regressions = []) {
   const criteriaMap = new Map(criteria.map((c) => [c.id, c]));
   let passed = 0, failed = 0, unsure = 0, blockerFailures = 0;
@@ -3019,9 +3045,20 @@ function registerEvalCommand(program) {
               `  Functional: ${outcomes.length}/${spec.test_cases.length} test case(s) executed`
             );
           }
+          const testCaseById = new Map(spec.test_cases.map((tc) => [tc.id, tc]));
           const allJudgments = [];
           for (const outcome of outcomes) {
-            const judgments = await judgeCriteria(spec.criteria, outcome, providers.judge, {
+            const testCase = testCaseById.get(outcome.test_case_id);
+            if (!testCase) {
+              throw new Error(
+                `Outcome references unknown test case id: "${outcome.test_case_id}"`
+              );
+            }
+            const applicableCriteria = selectCriteriaForTestCase(
+              spec.criteria,
+              testCase.criteria_ids
+            );
+            const judgments = await judgeCriteria(applicableCriteria, outcome, providers.judge, {
               model
             });
             for (const j of judgments) {
@@ -4207,7 +4244,7 @@ function registerSkillSignalCommands(program) {
 import { registerRefineCommand } from "@intentsolutions/refiner";
 function createProgram() {
   const program = new Command();
-  program.name("j-rig").description("Seven-layer binary evaluation harness for Claude Skills").version("0.1.0");
+  program.name("j-rig").description("Seven-layer binary evaluation harness for Claude Skills").version("0.1.1");
   registerCheckCommand(program);
   registerValidateCommand(program);
   registerEvalCommand(program);

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@intentsolutions/jrig-cli",
-  "version": "0.1.0",
+  "version": "0.1.1",
   "private": false,
   "description": "J-Rig seven-layer binary eval CLI for Claude Skills — the `j-rig` command: package integrity, trigger/functional/regression/baseline scoring, optimizer, and rollout-gate evidence. Self-contained (bundles the internal eval engine).",
   "keywords": [
@@ -61,8 +61,8 @@
     "zod": "^4.4.3"
   },
   "devDependencies": {
-    "@j-rig/core": "2.1.0",
     "@j-rig/db": "2.1.0",
+    "@j-rig/core": "2.1.0",
     "@j-rig/migrate": "2.1.0"
   },
   "scripts": {