@prismatic-io/lux 0.0.2-preview.18 → 0.0.2-preview.21

package/src/orchestrator/suite-compare.ts

@@ -5,6 +5,7 @@ import {
   MetricSummarySchema,
   type SuiteSample,
   type SuiteSummary,
+  suiteGroupKey,
   summarizeNumbers,
 } from "./suite-summary.js";
 
@@ -18,6 +19,7 @@ const ComparedOutcomeSchema = z
 export const SuiteCaseComparisonSchema = z
   .object({
     canonicalCaseId: z.string(),
+    permutationId: z.exactOptional(z.string()),
     before: ComparedOutcomeSchema.nullable(),
     after: ComparedOutcomeSchema.nullable(),
     passRateDelta: z.number().nullable(),
@@ -40,6 +42,7 @@ export const SuiteComparisonSchema = z
       z
         .object({
           canonicalCaseId: z.string(),
+          permutationId: z.exactOptional(z.string()),
           repetition: z.number().int().positive(),
           before: z.string().nullable(),
           after: z.string().nullable(),
@@ -54,9 +57,10 @@ export type SuiteComparison = z.infer<typeof SuiteComparisonSchema>;
 const samplesByCase = (summary: SuiteSummary): Map<string, SuiteSample[]> => {
   const out = new Map<string, SuiteSample[]>();
   for (const sample of summary.samples) {
-    const entries = out.get(sample.canonicalCaseId) ?? [];
+    const key = suiteGroupKey(sample);
+    const entries = out.get(key) ?? [];
     entries.push(sample);
-    out.set(sample.canonicalCaseId, entries);
+    out.set(key, entries);
   }
   return out;
 };
@@ -95,14 +99,23 @@ export const compareSuiteSummaries = (
   options: { passRateTolerance?: number } = {},
 ): SuiteComparison => {
   const passRateTolerance = options.passRateTolerance ?? 0;
-  const aCases = new Map(a.cases.map((entry) => [entry.canonicalCaseId, entry]));
-  const bCases = new Map(b.cases.map((entry) => [entry.canonicalCaseId, entry]));
+  const aCases = new Map(a.cases.map((entry) => [suiteGroupKey(entry), entry]));
+  const bCases = new Map(b.cases.map((entry) => [suiteGroupKey(entry), entry]));
   const aSamples = samplesByCase(a);
   const bSamples = samplesByCase(b);
-  const ids = new Set([...aCases.keys(), ...bCases.keys()]);
-  const identityMismatches = [...ids].sort().flatMap((canonicalCaseId) => {
-    const before = aSamples.get(canonicalCaseId) ?? [];
-    const after = bSamples.get(canonicalCaseId) ?? [];
+  const groups = new Map(
+    [...a.cases, ...b.cases].map((entry) => [
+      suiteGroupKey(entry),
+      {
+        canonicalCaseId: entry.canonicalCaseId,
+        ...(entry.permutationId ? { permutationId: entry.permutationId } : {}),
+      },
+    ]),
+  );
+  const keys = [...groups.keys()].sort();
+  const identityMismatches = keys.flatMap((key) => {
+    const before = aSamples.get(key) ?? [];
+    const after = bSamples.get(key) ?? [];
     return Array.from({ length: Math.max(before.length, after.length) }, (_, index) => {
       const beforeHash = before[index]?.identityHash ?? null;
       const afterHash = after[index]?.identityHash ?? null;
@@ -110,7 +123,7 @@ export const compareSuiteSummaries = (
         ? []
         : [
             {
-              canonicalCaseId,
+              ...groups.get(key),
               repetition: index + 1,
               before: beforeHash,
               after: afterHash,
@@ -118,9 +131,9 @@ export const compareSuiteSummaries = (
           ];
     }).flat();
   });
-  const cases = [...ids].sort().map((canonicalCaseId) => {
-    const before = aCases.get(canonicalCaseId);
-    const after = bCases.get(canonicalCaseId);
+  const cases = keys.map((key) => {
+    const before = aCases.get(key);
+    const after = bCases.get(key);
     let passRateDelta: number | null = null;
     if (before && after) {
       passRateDelta = after.passRate - before.passRate;
@@ -128,16 +141,13 @@ export const compareSuiteSummaries = (
       passRateDelta = -before.passRate;
     }
     return {
-      canonicalCaseId,
+      ...groups.get(key),
       before: before ? { attempts: before.attempts, passRate: before.passRate } : null,
       after: after ? { attempts: after.attempts, passRate: after.passRate } : null,
       passRateDelta,
       regressed:
         Boolean(before) && (!after || after.passRate + passRateTolerance < (before?.passRate ?? 0)),
-      pairedMetricDeltas: pairedMetricDeltas(
-        aSamples.get(canonicalCaseId) ?? [],
-        bSamples.get(canonicalCaseId) ?? [],
-      ),
+      pairedMetricDeltas: pairedMetricDeltas(aSamples.get(key) ?? [], bSamples.get(key) ?? []),
     };
   });
   return SuiteComparisonSchema.parse({
@@ -182,13 +192,13 @@ export const formatSuiteComparisonText = (comparison: SuiteComparison): string =
   if (comparison.identityMismatches.length > 0) {
     lines.push(`  identity mismatches: ${comparison.identityMismatches.length}`);
     for (const mismatch of comparison.identityMismatches) {
-      lines.push(`    ! ${mismatch.canonicalCaseId} repetition ${mismatch.repetition}`);
+      lines.push(`    ! ${suiteGroupKey(mismatch)} repetition ${mismatch.repetition}`);
     }
   }
   for (const entry of comparison.cases) {
     if (!entry.regressed) continue;
     lines.push(
-      `    ↓ ${entry.canonicalCaseId}: ${entry.before ? percent(entry.before.passRate) : "—"} → ${entry.after ? percent(entry.after.passRate) : "missing"}`,
+      `    ↓ ${suiteGroupKey(entry)}: ${entry.before ? percent(entry.before.passRate) : "—"} → ${entry.after ? percent(entry.after.passRate) : "missing"}`,
     );
   }
   return lines.join("\n");

package/src/orchestrator/suite-runs.ts

@@ -10,6 +10,7 @@ const SUITE_FILE = "suite.json";
 const SuiteItemSchema = z.object({
   caseId: z.string(),
   canonicalCaseId: z.string(),
+  permutationId: z.exactOptional(z.string()),
   status: z.enum(["pending", "running", "completed"]),
   runDir: z.string().nullable(),
   exitReason: z.union([z.literal("done"), z.string()]).nullable(),
@@ -64,7 +65,7 @@ export type OpenedSuiteRun = {
 export const openSuiteRun = async (options: {
   runsRoot: string;
   concurrency: number;
-  items: { id: string; caseId?: string | undefined }[];
+  items: { id: string; caseId?: string | undefined; permutationId?: string | undefined }[];
   startedAt?: Date | undefined;
   profileId?: string | undefined;
   experimentId?: string | undefined;
@@ -110,6 +111,7 @@ export const openSuiteRun = async (options: {
     items: options.items.map((item) => ({
       caseId: item.id,
       canonicalCaseId: item.caseId ?? item.id,
+      ...(item.permutationId ? { permutationId: item.permutationId } : {}),
       status: "pending",
       runDir: null,
       exitReason: null,

package/src/orchestrator/suite-summary.ts

@@ -42,14 +42,25 @@ export type OutcomeSummary = z.infer<typeof OutcomeSummarySchema>;
 
 const CaseSummarySchema = OutcomeSummarySchema.extend({
   canonicalCaseId: z.string(),
+  permutationId: z.exactOptional(z.string()),
   metrics: z.record(z.string(), MetricSummarySchema),
 });
+
+/** Repetitions pool per case and harness permutation, never across permutations. */
+export const suiteGroupKey = (entry: {
+  canonicalCaseId: string;
+  permutationId?: string | undefined;
+}): string =>
+  entry.permutationId === undefined
+    ? entry.canonicalCaseId
+    : `${entry.canonicalCaseId} [${entry.permutationId}]`;
 export type CaseSummary = z.infer<typeof CaseSummarySchema>;
 
 export const SuiteSampleSchema = z.object({
   index: z.number().int().nonnegative(),
   caseId: z.string(),
   canonicalCaseId: z.string(),
+  permutationId: z.exactOptional(z.string()),
   exitReason: z.string().nullable(),
   passed: z.boolean().nullable(),
   identityHash: z.string().nullable(),
@@ -308,14 +319,15 @@ export const buildSuiteSummary = (
 
   const byCase = new Map<string, typeof entries>();
   for (const entry of entries) {
-    const group = byCase.get(entry.item.canonicalCaseId) ?? [];
+    const key = suiteGroupKey(entry.item);
+    const group = byCase.get(key) ?? [];
     group.push(entry);
-    byCase.set(entry.item.canonicalCaseId, group);
+    byCase.set(key, group);
   }
 
   const cases = [...byCase.entries()]
     .sort(([a], [b]) => a.localeCompare(b))
-    .map(([canonicalCaseId, caseEntries]) => {
+    .map(([, caseEntries]) => {
       const metrics: MetricSamples = new Map();
       for (const entry of caseEntries) {
         const result = entry.result?.result;
@@ -323,8 +335,10 @@ export const buildSuiteSummary = (
         add(metrics, "score", result.grading.caseScore ?? undefined);
         collectMetadata(metrics, result.metadata);
       }
+      const first = caseEntries[0]?.item;
       return {
-        canonicalCaseId,
+        canonicalCaseId: first?.canonicalCaseId ?? "",
+        ...(first?.permutationId ? { permutationId: first.permutationId } : {}),
         ...summarizeOutcomes(caseEntries),
         metrics: summarizeMetrics(metrics),
       };
@@ -345,6 +359,7 @@ export const buildSuiteSummary = (
         index,
         caseId: entry.item.caseId,
         canonicalCaseId: entry.item.canonicalCaseId,
+        ...(entry.item.permutationId ? { permutationId: entry.item.permutationId } : {}),
         exitReason: entry.result?.exitReason ?? null,
         passed: entry.result?.casePassed ?? null,
         identityHash: result ? runComparisonIdentity(result.metadata) : null,

package/lib/drivers/cursor/acp-transport.d.ts.map

@@ -1 +0,0 @@
-{"version":3,"file":"acp-transport.d.ts","sourceRoot":"","sources":["../../../src/drivers/cursor/acp-transport.ts"],"names":[],"mappings":"AAEA,MAAM,MAAM,SAAS,GAAG,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,CAAC;AAChD,MAAM,MAAM,KAAK,GAAG,MAAM,GAAG,MAAM,CAAC;AACpC,eAAO,MAAM,QAAQ,UAAW,OAAO,KAAG,SAAS,GAAG,SAGvC,CAAC;AAEhB,KAAK,gBAAgB,GAAG;IACtB,OAAO,EAAE,MAAM,CAAC;IAChB,IAAI,EAAE,MAAM,EAAE,CAAC;IACf,GAAG,EAAE,MAAM,CAAC;IACZ,GAAG,CAAC,EAAE,MAAM,CAAC,UAAU,CAAC;IACxB,MAAM,EAAE,WAAW,CAAC;IACpB,YAAY,EAAE,MAAM,CAAC;IACrB,cAAc,EAAE,MAAM,CAAC;IACvB,SAAS,EAAE,CAAC,OAAO,EAAE,SAAS,KAAK,OAAO,CAAC,IAAI,CAAC,CAAC;IACjD,SAAS,EAAE,CAAC,KAAK,EAAE,KAAK,KAAK,IAAI,CAAC;IAClC,UAAU,EAAE,MAAM,IAAI,CAAC;CACxB,CAAC;AAEF,qBAAa,YAAY;IACvB,OAAO,CAAC,QAAQ,CAAC,OAAO,CAAiB;IACzC,OAAO,CAAC,QAAQ,CAAC,OAAO,CAOpB;IACJ,OAAO,CAAC,MAAM,CAAK;IACnB,OAAO,CAAC,MAAM,CAAmB;IACjC,OAAO,CAAC,WAAW,CAAK;IACxB,OAAO,CAAC,KAAK,CAAoC;IACjD,OAAO,CAAC,OAAO,CAAS;IACxB,OAAO,CAAC,UAAU,CAAM;IAExB,OAAO,CAAC,QAAQ,CAAC,OAAO,CAAmB;IAE3C,YAAY,OAAO,EAAE,gBAAgB,EAoBpC;IAEK,OAAO,IAAI,OAAO,CAAC,IAAI,CAAC,CAE7B;IAED,OAAO,CAAC,MAAM,EAAE,MAAM,EAAE,MAAM,EAAE,SAAS,EAAE,SAAS,CAAC,EAAE,MAAM,GAAG,OAAO,CAAC,SAAS,CAAC,CAiBjF;IAED,KAAK,CAAC,OAAO,EAAE,SAAS,GAAG,IAAI,CAM9B;IAEK,OAAO,IAAI,OAAO,CAAC,IAAI,CAAC,CAE7B;IAEK,KAAK,IAAI,OAAO,CAAC,IAAI,CAAC,CAK3B;IAED,OAAO,CAAC,aAAa;IAQrB,OAAO,CAAC,IAAI;IAQZ,OAAO,CAAC,IAAI;IAmBZ,OAAO,CAAC,IAAI;CAyCb"}

package/lib/drivers/cursor/acp-transport.js.map

@@ -1 +0,0 @@
-{"version":3,"file":"acp-transport.js","sourceRoot":"","sources":["../../../src/drivers/cursor/acp-transport.ts"],"names":[],"mappings":"AAAA,OAAO,EAAuB,mBAAmB,EAAE,MAAM,qBAAqB,CAAC;AAI/E,MAAM,CAAC,MAAM,QAAQ,GAAG,CAAC,KAAc,EAAyB,EAAE,CAChE,KAAK,KAAK,IAAI,IAAI,OAAO,KAAK,KAAK,QAAQ,IAAI,CAAC,KAAK,CAAC,OAAO,CAAC,KAAK,CAAC;IAClE,CAAC,CAAE,KAAmB;IACtB,CAAC,CAAC,SAAS,CAAC;AAehB,MAAM,OAAO,YAAY;IACN,OAAO,CAAiB;IACxB,OAAO,GAAG,IAAI,GAAG,EAO/B,CAAC;IACI,MAAM,GAAG,CAAC,CAAC;IACX,MAAM,GAAG,MAAM,CAAC,KAAK,CAAC,CAAC,CAAC,CAAC;IACzB,WAAW,GAAG,CAAC,CAAC;IAChB,KAAK,GAAkB,OAAO,CAAC,OAAO,EAAE,CAAC;IACzC,OAAO,GAAG,KAAK,CAAC;IAChB,UAAU,GAAG,EAAE,CAAC;IAEP,OAAO,CAAmB;IAE3C,YAAY,OAAyB;QACnC,IAAI,CAAC,OAAO,GAAG,OAAO,CAAC;QACvB,IAAI,CAAC,OAAO,GAAG,mBAAmB,CAAC,OAAO,CAAC,OAAO,EAAE,OAAO,CAAC,IAAI,EAAE;YAChE,GAAG,EAAE,OAAO,CAAC,GAAG;YAChB,GAAG,CAAC,OAAO,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,GAAG,EAAE,OAAO,CAAC,GAAG,EAAE,CAAC,CAAC,CAAC,EAAE,CAAC;YAC5C,MAAM,EAAE,OAAO,CAAC,MAAM;SACvB,CAAC,CAAC;QACH,IAAI,CAAC,OAAO,CAAC,KAAK,CAAC,MAAM,CAAC,EAAE,CAAC,MAAM,EAAE,CAAC,KAAa,EAAE,EAAE,CAAC,IAAI,CAAC,IAAI,CAAC,KAAK,CAAC,CAAC,CAAC;QAC1E,IAAI,CAAC,OAAO,CAAC,KAAK,CAAC,MAAM,CAAC,WAAW,CAAC,MAAM,CAAC,CAAC;QAC9C,IAAI,CAAC,OAAO,CAAC,KAAK,CAAC,MAAM,CAAC,EAAE,CAAC,MAAM,EAAE,CAAC,KAAa,EAAE,EAAE;YACrD,IAAI,CAAC,UAAU,GAAG,CAAC,IAAI,CAAC,UAAU,GAAG,KAAK,CAAC,CAAC,KAAK,CAAC,CAAC,KAAK,CAAC,CAAC;QAC5D,CAAC,CAAC,CAAC;QACH,KAAK,IAAI,CAAC,OAAO,CAAC,MAAM,CAAC,IAAI,CAAC,CAAC,EAAE,QAAQ,EAAE,MAAM,EAAE,EAAE,EAAE;YACrD,IAAI,CAAC,IAAI,CAAC,OAAO;gBACf,IAAI,CAAC,IAAI,CACP,IAAI,KAAK,CACP,6CAA6C,QAAQ,YAAY,MAAM,IAAI,IAAI,CAAC,UAAU,CAAC,CAAC,CAAC,KAAK,IAAI,CAAC,UAAU,EAAE,CAAC,CAAC,CAAC,EAAE,EAAE,CAC3H,CACF,CAAC;QACN,CAAC,CAAC,CAAC;IACL,CAAC;IAED,KAAK,CAAC,OAAO;QACX,MAAM,IAAI,CAAC,OAAO,CAAC,OAAO,CAAC;IAC7B,CAAC;IAED,OAAO,CAAC,MAAc,EAAE,MAAiB,EAAE,SAAkB;QAC3D,IAAI,IAAI,CAAC,OAAO;YAAE,OAAO,OAAO,CAAC,MAAM,CAAC,IAAI,KAAK,CAAC,sBAAsB,CAAC,CAAC,CAAC;QAC3E,MAAM,EAAE,GAAG,IAAI,CAAC,MAAM,EAAE,CAAC;QACzB,OAAO,IAAI,OAAO,CAAC,CAAC,OAAO,EAAE,MAAM,EAAE,EAAE;YACrC,MAAM,KAAK,GACT,SAAS,KAAK,SAAS;gBACrB,CAAC,CAAC,SAAS;gBACX,CAAC,CAAC,UAAU,CAAC,GAAG,EAAE;oBACd,IAAI,CAAC,IAAI,CAAC,IAAI,KAAK,CAAC,cAAc,MAAM,oBAAoB,SAAS,IAAI,CAAC,CAAC,CAAC;gBAC9E,CAAC,EAAE,SAAS,CAAC,CAAC;YACpB,IAAI,CAAC,OAAO,CAAC,GAAG,CAAC,EAAE,EAAE,EAAE,OAAO,EAAE,MAAM,EAAE,KAAK,EAAE,CAAC,CAAC;YACjD,IAAI,CAAC;gBACH,IAAI,CAAC,KAAK,CAAC,EAAE,EAAE,EAAE,MAAM,EAAE,MAAM,EAAE,CAAC,CAAC;YACrC,CAAC;YAAC,OAAO,KAAK,EAAE,CAAC;gBACf,IAAI,CAAC,IAAI,CAAC,KAAK,YAAY,KAAK,CAAC,CAAC,CAAC,KAAK,CAAC,CAAC,CAAC,IAAI,KAAK,CAAC,MAAM,CAAC,KAAK,CAAC,CAAC,CAAC,CAAC;YACvE,CAAC;QACH,CAAC,CAAC,CAAC;IACL,CAAC;IAED,KAAK,CAAC,OAAkB;QACtB,IAAI,IAAI,CAAC,OAAO;YAAE,MAAM,IAAI,KAAK,CAAC,sBAAsB,CAAC,CAAC;QAC1D,MAAM,IAAI,GAAG,GAAG,IAAI,CAAC,SAAS,CAAC,EAAE,OAAO,EAAE,KAAK,EAAE,GAAG,OAAO,EAAE,CAAC,IAAI,CAAC;QACnE,IAAI,MAAM,CAAC,UAAU,CAAC,IAAI,CAAC,GAAG,IAAI,CAAC,OAAO,CAAC,YAAY;YACrD,MAAM,IAAI,KAAK,CAAC,mDAAmD,CAAC,CAAC;QACvE,IAAI,CAAC,OAAO,CAAC,KAAK,CAAC,KAAK,CAAC,KAAK,CAAC,IAAI,CAAC,CAAC;IACvC,CAAC;IAED,KAAK,CAAC,OAAO;QACX,MAAM,IAAI,CAAC,KAAK,CAAC;IACnB,CAAC;IAED,KAAK,CAAC,KAAK;QACT,IAAI,CAAC,OAAO,GAAG,IAAI,CAAC;QACpB,IAAI,CAAC,aAAa,CAAC,IAAI,KAAK,CAAC,sBAAsB,CAAC,CAAC,CAAC;QACtD,IAAI,CAAC,MAAM,GAAG,MAAM,CAAC,KAAK,CAAC,CAAC,CAAC,CAAC;QAC9B,MAAM,IAAI,CAAC,OAAO,CAAC,SAAS,EAAE,CAAC;IACjC,CAAC;IAEO,aAAa,CAAC,KAAY;QAChC,KAAK,MAAM,OAAO,IAAI,IAAI,CAAC,OAAO,CAAC,MAAM,EAAE,EAAE,CAAC;YAC5C,YAAY,CAAC,OAAO,CAAC,KAAK,CAAC,CAAC;YAC5B,OAAO,CAAC,MAAM,CAAC,KAAK,CAAC,CAAC;QACxB,CAAC;QACD,IAAI,CAAC,OAAO,CAAC,KAAK,EAAE,CAAC;IACvB,CAAC;IAEO,IAAI,CAAC,KAAY;QACvB,IAAI,IAAI,CAAC,OAAO;YAAE,OAAO;QACzB,IAAI,CAAC,OAAO,GAAG,IAAI,CAAC;QACpB,IAAI,CAAC,aAAa,CAAC,KAAK,CAAC,CAAC;QAC1B,IAAI,CAAC,OAAO,CAAC,SAAS,CAAC,KAAK,CAAC,CAAC;QAC9B,KAAK,IAAI,CAAC,OAAO,CAAC,SAAS,EAAE,CAAC;IAChC,CAAC;IAEO,IAAI,CAAC,KAAa;QACxB,IAAI,MAAM,GAAG,CAAC,CAAC;QACf,OAAO,MAAM,GAAG,KAAK,CAAC,MAAM,IAAI,CAAC,IAAI,CAAC,OAAO,EAAE,CAAC;YAC9C,MAAM,OAAO,GAAG,KAAK,CAAC,OAAO,CAAC,IAAI,EAAE,MAAM,CAAC,CAAC;YAC5C,MAAM,GAAG,GAAG,OAAO,KAAK,CAAC,CAAC,CAAC,CAAC,CAAC,KAAK,CAAC,MAAM,CAAC,CAAC,CAAC,OAAO,CAAC;YACpD,MAAM,OAAO,GAAG,KAAK,CAAC,QAAQ,CAAC,MAAM,EAAE,GAAG,CAAC,CAAC;YAC5C,IAAI,IAAI,CAAC,MAAM,CAAC,MAAM,GAAG,OAAO,CAAC,MAAM,GAAG,IAAI,CAAC,OAAO,CAAC,YAAY,EAAE,CAAC;gBACpE,IAAI,CAAC,IAAI,CAAC,IAAI,KAAK,CAAC,8CAA8C,CAAC,CAAC,CAAC;gBACrE,OAAO;YACT,CAAC;YACD,IAAI,CAAC,MAAM,GAAG,MAAM,CAAC,MAAM,CAAC,CAAC,IAAI,CAAC,MAAM,EAAE,OAAO,CAAC,CAAC,CAAC;YACpD,IAAI,OAAO,KAAK,CAAC,CAAC;gBAAE,OAAO;YAC3B,MAAM,IAAI,GAAG,IAAI,CAAC,MAAM,CAAC;YACzB,IAAI,CAAC,MAAM,GAAG,MAAM,CAAC,KAAK,CAAC,CAAC,CAAC,CAAC;YAC9B,IAAI,CAAC,IAAI,CAAC,IAAI,CAAC,CAAC;YAChB,MAAM,GAAG,OAAO,GAAG,CAAC,CAAC;QACvB,CAAC;IACH,CAAC;IAEO,IAAI,CAAC,IAAY;QACvB,IAAI,CAAC,IAAI,CAAC,QAAQ,CAAC,MAAM,CAAC,CAAC,IAAI,EAAE;YAAE,OAAO;QAC1C,IAAI,OAA8B,CAAC;QACnC,IAAI,CAAC;YACH,OAAO,GAAG,QAAQ,CAAC,IAAI,CAAC,KAAK,CAAC,IAAI,CAAC,QAAQ,CAAC,MAAM,CAAC,CAAC,CAAC,CAAC;QACxD,CAAC;QAAC,MAAM,CAAC;YACP,qBAAqB;QACvB,CAAC;QACD,IAAI,OAAO,EAAE,OAAO,KAAK,KAAK,EAAE,CAAC;YAC/B,IAAI,CAAC,IAAI,CAAC,IAAI,KAAK,CAAC,iDAAiD,CAAC,CAAC,CAAC;YACxE,OAAO;QACT,CAAC;QACD,IAAI,CAAC,OAAO,CAAC,UAAU,EAAE,CAAC;QAC1B,IAAI,OAAO,CAAC,MAAM,KAAK,SAAS,EAAE,CAAC;YACjC,MAAM,OAAO,GAAG,IAAI,CAAC,OAAO,CAAC,GAAG,CAAC,OAAO,CAAC,EAAW,CAAC,CAAC;YACtD,IAAI,CAAC,OAAO;gBAAE,OAAO;YACrB,IAAI,CAAC,OAAO,CAAC,MAAM,CAAC,OAAO,CAAC,EAAW,CAAC,CAAC;YACzC,YAAY,CAAC,OAAO,CAAC,KAAK,CAAC,CAAC;YAC5B,MAAM,KAAK,GAAG,QAAQ,CAAC,OAAO,CAAC,KAAK,CAAC,CAAC;YACtC,IAAI,KAAK;gBAAE,OAAO,CAAC,MAAM,CAAC,IAAI,KAAK,CAAC,MAAM,CAAC,KAAK,CAAC,OAAO,IAAI,2BAA2B,CAAC,CAAC,CAAC,CAAC;;gBACtF,OAAO,CAAC,OAAO,CAAC,QAAQ,CAAC,OAAO,CAAC,MAAM,CAAC,IAAI,EAAE,CAAC,CAAC;YACrD,OAAO;QACT,CAAC;QACD,IAAI,CAAC,WAAW,IAAI,IAAI,CAAC,MAAM,CAAC;QAChC,IAAI,IAAI,CAAC,WAAW,GAAG,IAAI,CAAC,OAAO,CAAC,cAAc,EAAE,CAAC;YACnD,IAAI,CAAC,IAAI,CAAC,IAAI,KAAK,CAAC,mDAAmD,CAAC,CAAC,CAAC;YAC1E,OAAO;QACT,CAAC;QACD,MAAM,MAAM,GAAG,OAAO,CAAC;QACvB,IAAI,CAAC,KAAK,GAAG,IAAI,CAAC,KAAK;aACpB,IAAI,CAAC,KAAK,IAAI,EAAE;YACf,IAAI,CAAC;gBACH,IAAI,CAAC,IAAI,CAAC,OAAO;oBAAE,MAAM,IAAI,CAAC,OAAO,CAAC,SAAS,CAAC,MAAM,CAAC,CAAC;YAC1D,CAAC;oBAAS,CAAC;gBACT,IAAI,CAAC,WAAW,IAAI,IAAI,CAAC,MAAM,CAAC;YAClC,CAAC;QACH,CAAC,CAAC;aACD,KAAK,CAAC,CAAC,KAAc,EAAE,EAAE,CACxB,IAAI,CAAC,IAAI,CAAC,KAAK,YAAY,KAAK,CAAC,CAAC,CAAC,KAAK,CAAC,CAAC,CAAC,IAAI,KAAK,CAAC,MAAM,CAAC,KAAK,CAAC,CAAC,CAAC,CACrE,CAAC;IACN,CAAC;CACF"}

package/lib/drivers/cursor/events.d.ts.map

@@ -1 +0,0 @@
-{"version":3,"file":"events.d.ts","sourceRoot":"","sources":["../../../src/drivers/cursor/events.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,KAAK,QAAQ,EAAsB,MAAM,qBAAqB,CAAC;AACxE,OAAO,EAAY,KAAK,SAAS,EAAE,MAAM,oBAAoB,CAAC;AAK9D,qBAAa,YAAY;IACvB,OAAO,CAAC,GAAG,CAAK;IAChB,OAAO,CAAC,aAAa,CAAK;IAC1B,OAAO,CAAC,QAAQ,CAAC,KAAK,CAAgC;IACtD,OAAO,CAAC,QAAQ,CAAC,SAAS,CAAqB;IAE/C,OAAO,CAAC,QAAQ,CAAC,YAAY,CAAS;IACtC,OAAO,CAAC,QAAQ,CAAC,QAAQ,CAAS;IAElC,YAAY,YAAY,EAAE,MAAM,EAAE,QAAQ,EAAE,MAAM,EAGjD;IAED,MAAM,CAAC,MAAM,EAAE,SAAS,GAAG,QAAQ,EAAE,CA6DpC;CACF"}

package/lib/drivers/cursor/events.js.map

@@ -1 +0,0 @@
-{"version":3,"file":"events.js","sourceRoot":"","sources":["../../../src/drivers/cursor/events.ts"],"names":[],"mappings":"AAAA,OAAO,EAAiB,kBAAkB,EAAE,MAAM,qBAAqB,CAAC;AACxE,OAAO,EAAE,QAAQ,EAAkB,MAAM,oBAAoB,CAAC;AAC9D,OAAO,EAAE,OAAO,EAAE,MAAM,kBAAkB,CAAC;AAI3C,MAAM,OAAO,YAAY;IACf,GAAG,GAAG,CAAC,CAAC;IACR,aAAa,GAAG,CAAC,CAAC;IACT,KAAK,GAAG,IAAI,GAAG,EAAqB,CAAC;IACrC,SAAS,GAAG,IAAI,GAAG,EAAU,CAAC;IAE9B,YAAY,CAAS;IACrB,QAAQ,CAAS;IAElC,YAAY,YAAoB,EAAE,QAAgB;QAChD,IAAI,CAAC,YAAY,GAAG,YAAY,CAAC;QACjC,IAAI,CAAC,QAAQ,GAAG,QAAQ,CAAC;IAC3B,CAAC;IAED,MAAM,CAAC,MAAiB;QACtB,MAAM,EAAE,GAAG,IAAI,CAAC,GAAG,EAAE,CAAC;QACtB,IACE,MAAM,CAAC,aAAa,KAAK,qBAAqB;YAC9C,MAAM,CAAC,aAAa,KAAK,qBAAqB,EAC9C,CAAC;YACD,MAAM,OAAO,GAAG,QAAQ,CAAC,MAAM,CAAC,OAAO,CAAC,CAAC;YACzC,IAAI,OAAO,EAAE,IAAI,KAAK,MAAM,IAAI,OAAO,OAAO,CAAC,IAAI,KAAK,QAAQ;gBAAE,OAAO,EAAE,CAAC;YAC5E,OAAO;gBACL;oBACE,EAAE,EAAE,kBAAkB,EAAE,IAAI,CAAC,GAAG,EAAE;oBAClC,IAAI,EAAE,MAAM,CAAC,aAAa,KAAK,qBAAqB,CAAC,CAAC,CAAC,eAAe,CAAC,CAAC,CAAC,WAAW;oBACpF,OAAO,EAAE,EAAE,IAAI,EAAE,OAAO,CAAC,IAAI,EAAE;oBAC/B,EAAE;iBACH;aACF,CAAC;QACJ,CAAC;QACD,IAAI,MAAM,CAAC,aAAa,KAAK,WAAW,IAAI,MAAM,CAAC,aAAa,KAAK,kBAAkB,EAAE,CAAC;YACxF,OAAO,CAAC,EAAE,EAAE,EAAE,iBAAiB,EAAE,IAAI,CAAC,GAAG,EAAE,EAAE,IAAI,EAAE,eAAe,EAAE,OAAO,EAAE,MAAM,EAAE,EAAE,EAAE,CAAC,CAAC;QAC7F,CAAC;QACD,MAAM,EAAE,GAAG,MAAM,CAAC,UAAU,CAAC;QAC7B,IAAI,OAAO,EAAE,KAAK,QAAQ;YAAE,MAAM,IAAI,KAAK,CAAC,2CAA2C,CAAC,CAAC;QACzF,IAAI,IAAI,CAAC,SAAS,CAAC,GAAG,CAAC,EAAE,CAAC;YAAE,OAAO,EAAE,CAAC;QACtC,MAAM,GAAG,GAAG,IAAI,CAAC,KAAK,CAAC,GAAG,CAAC,EAAE,CAAC,CAAC;QAC/B,IAAI,CAAC,GAAG,IAAI,IAAI,CAAC,KAAK,CAAC,IAAI,GAAG,IAAI,CAAC,SAAS,CAAC,IAAI,IAAI,IAAI,CAAC,YAAY;YACpE,MAAM,IAAI,KAAK,CAAC,kCAAkC,CAAC,CAAC;QACtD,MAAM,KAAK,GAAG,EAAE,GAAG,GAAG,EAAE,KAAK,EAAE,GAAG,MAAM,EAAE,CAAC;QAC3C,MAAM,KAAK,GAAG,MAAM,CAAC,UAAU,CAAC,IAAI,CAAC,SAAS,CAAC,KAAK,CAAC,CAAC,CAAC;QACvD,IAAI,CAAC,aAAa,IAAI,KAAK,GAAG,CAAC,GAAG,EAAE,KAAK,IAAI,CAAC,CAAC,CAAC;QAChD,IAAI,IAAI,CAAC,aAAa,GAAG,IAAI,CAAC,QAAQ;YACpC,MAAM,IAAI,KAAK,CAAC,+CAA+C,CAAC,CAAC;QACnE,MAAM,IAAI,GAAG,GAAG,EAAE,IAAI,IAAI,MAAM,CAAC,KAAK,CAAC,KAAK,IAAI,KAAK,CAAC,IAAI,IAAI,aAAa,CAAC,CAAC;QAC7E,IAAI,CAAC,KAAK,CAAC,GAAG,CAAC,EAAE,EAAE,EAAE,KAAK,EAAE,KAAK,EAAE,IAAI,EAAE,CAAC,CAAC;QAC3C,MAAM,MAAM,GAAe,GAAG;YAC5B,CAAC,CAAC,EAAE;YACJ,CAAC,CAAC;gBACE;oBACE,EAAE,EAAE,MAAM,EAAE,EAAE;oBACd,IAAI,EAAE,WAAW;oBACjB,OAAO,EAAE,EAAE,IAAI,EAAE,KAAK,EAAE,KAAK,CAAC,QAAQ,IAAI,EAAE,EAAE,SAAS,EAAE,EAAE,EAAE;oBAC7D,EAAE;iBACH;aACF,CAAC;QACN,IAAI,KAAK,CAAC,MAAM,KAAK,WAAW,IAAI,KAAK,CAAC,MAAM,KAAK,QAAQ,EAAE,CAAC;YAC9D,MAAM,OAAO,GAAG,KAAK,CAAC,SAAS,IAAI,KAAK,CAAC,OAAO,IAAI,EAAE,CAAC;YACvD,MAAM,IAAI,GAAG,OAAO,CAAC,KAAK,CAAC,OAAO,CAAC;iBAChC,GAAG,CAAC,CAAC,KAAK,EAAE,EAAE,CAAC,QAAQ,CAAC,KAAK,CAAC,OAAO,CAAC,EAAE,IAAI,CAAC;iBAC7C,MAAM,CAAC,CAAC,KAAK,EAAE,EAAE,CAAC,OAAO,KAAK,KAAK,QAAQ,CAAC;iBAC5C,IAAI,CAAC,IAAI,CAAC,CAAC;YACd,MAAM,CAAC,IAAI,CACT,kBAAkB,CAChB,EAAE,IAAI,CAAC,GAAG,EACV,EAAE,IAAI,EAAE,SAAS,EAAE,EAAE,EAAE,OAAO,EAAE,OAAO,EAAE,KAAK,CAAC,MAAM,KAAK,QAAQ,EAAE,IAAI,EAAE,EAC1E,EAAE,CACH,CACF,CAAC;YACF,IAAI,CAAC,aAAa,IAAI,KAAK,CAAC;YAC5B,IAAI,CAAC,KAAK,CAAC,MAAM,CAAC,EAAE,CAAC,CAAC;YACtB,IAAI,CAAC,SAAS,CAAC,GAAG,CAAC,EAAE,CAAC,CAAC;QACzB,CAAC;QACD,OAAO,MAAM,CAAC;IAChB,CAAC;CACF"}

package/skills/lux/SKILL.md

@@ -1,89 +0,0 @@
----
-name: lux
-description: Author, validate, run, inspect, compare, and improve coding-agent evaluations with the Lux CLI. Use when creating or reviewing Lux cases, custom assertions, lifecycle fixtures, driver or harness configuration, campaigns, grader annotations, diagnosing eval failures, comparing runs, or running and promoting Lux experiments.
----
-
-# Lux
-
-Use Lux as an eval test runner and evidence system for coding agents. Prefer
-small, discriminating cases with deterministic assertions; add an LLM rubric
-only for qualities that artifacts and commands cannot establish.
-
-## Start here
-
-1. Inspect the repository and its existing `lux.config.ts`, cases, fixtures,
-   prompts, and experiments before adding files.
-2. If Lux is not initialized, run `npm exec -- lux init` for ordinary evals or
-   `npm exec -- lux init --experiment` for an improvement campaign.
-3. Read [references/eval-authoring.md](references/eval-authoring.md) before
-   authoring or reviewing a case, assertion plugin, or lifecycle fixture.
-4. Read [references/cli.md](references/cli.md) before running, diagnosing,
-   comparing, or promoting results.
-5. Run `npm exec -- lux doctor` before model calls. Treat its errors as
-   authoring problems, not as permission to bypass validation.
-
-## Workflow
-
-### Author an eval
-
-- Define the behavior and likely failure mode first.
-- Make the prompt realistic and self-contained.
-- Give the persona only facts and preferences the simulated user should know.
-- Assert observable outcomes, not implementation details.
-- Use stable assertion IDs so grading evidence remains intelligible.
-- Use the cheapest deterministic assertion that proves each requirement.
-- Include a rubric assertion only when semantic judgment is unavoidable, and
-  write one narrow, falsifiable criterion per assertion.
-- Pin the subject driver model explicitly when comparability matters.
-
-### Run and diagnose
-
-```bash
-npm exec -- lux doctor
-npm exec -- lux run <case-or-filter> --list
-npm exec -- lux run <case-or-filter> --loop 3
-npm exec -- lux view <run-or-suite>
-```
-
-Inspect `case.json`, `run.json`, `events.jsonl`, `grading.json`, and `artifacts/`
-before changing the case. Separate subject failure from harness failure:
-
-- Subject failure: the agent misunderstood, asked poorly, used the wrong tools,
-  or left incorrect artifacts.
-- Harness failure: the prompt was underspecified, the persona lacked a needed
-  fact, a fixture or driver failed, an assertion measured a proxy, or the
-  judge criterion was ambiguous.
-
-Do not weaken a valid assertion merely to make a run pass.
-
-### Improve with evidence
-
-Use repeated runs and held-out splits for noisy agent behavior. For skill,
-prompt, or agent-source changes, use `subjectPath()` in cases and an experiment
-campaign with immutable candidates. Plan first, run the experiment, inspect the
-report, and apply a promoted candidate as a separate action:
-
-```bash
-npm exec -- lux experiment experiments/example.experiment.ts --plan
-npm exec -- lux experiment experiments/example.experiment.ts
-npm exec -- lux report <experiment-dir>
-npm exec -- lux apply <experiment-dir>
-```
-
-Never edit a materialized candidate or managed experiment record by hand.
-
-## Guardrails
-
-- Lux eval projects execute project-authored commands and agent-generated work.
-  Run untrusted evals only in a disposable environment.
-- Do not infer success from agent prose. Grade the final artifacts and recorded
-  behavior.
-- Do not use only `run-succeeded` for a case that claims to verify a concrete
-  result.
-- Do not invoke bare `lux`; it displays help. Use the explicit `lux run` verb
-  and narrow by case, path, or tag before any model call.
-- Do not let train examples leak into validation or test splits.
-- Do not combine evidence across changed models, harnesses, cases, or subjects;
-  Lux identity checks exist to prevent that.
-- Prefer `--plan`, `doctor`, `view`, and `report` before expensive or mutating
-  commands.

package/skills/lux/references/cli.md

@@ -1,128 +0,0 @@
-# Lux CLI runbook
-
-Invoke the project-local CLI with `npm exec -- lux ...` unless the repository
-uses another package-manager convention.
-
-Bare `lux` prints command and agent-skill help. Running cases always requires
-the explicit `run` verb; do not assume an unfiltered run is harmless.
-Use `lux <command> --help` before composing an unfamiliar or automation-facing
-invocation; it is the authoritative option list for the installed version.
-Use `lux --llms` for the compact command manifest and
-`lux <command> --schema --json` for exact argument, option, and output schemas.
-All commands support Incur's `--format`, `--filter-output`, and token-pagination
-flags; prefer the default TOON output unless strict JSON is required.
-
-## Author and preflight
-
-```bash
-lux init
-lux init --experiment
-lux skills list
-lux skills add --no-global
-lux doctor
-lux doctor experiments/campaign.experiment.ts --json
-lux run --list
-lux run --list-tags
-```
-
-`init` is conservative around existing files. `doctor` validates configuration,
-case discovery, driver executables, campaign splits, confidence feasibility,
-and projected budgets without model calls.
-
-## Run
-
-```bash
-lux run
-lux run retry-flag
-lux run cases/retry-flag.ts
-lux run --tag smoke
-lux run retry --loop 5 --concurrency 2
-lux run one-case --interactive
-lux run one-case --claude-answerer
-lux run one-case --profile small-low --subject-root . --verbose
-lux run one-case --skip-judge
-```
-
-Filters match case IDs and paths. Tags come from `meta.tags`. A run invocation
-also creates a suite record, which is the right unit for repeated or multi-case
-comparison.
-
-`--skip-judge` leaves subject execution unchanged but does not call rubric LLM
-judges. Deterministic assertions still determine the verdict; judge-only cases
-remain ungraded, and persisted grading records identify the skipped rubrics.
-
-## Inspect and compare
-
-```bash
-lux view
-lux view <run-or-suite>
-lux view --suites --json
-lux view --experiments
-lux compare <run-or-suite-a> <run-or-suite-b> --pass-rate-tolerance 0.05
-```
-
-Use `view` before opening raw files. When diagnosing, inspect:
-
-- `case.json`: frozen authored inputs;
-- `run.json`: exit reason, timing, usage, source identity;
-- `events.jsonl`: trajectory, tools, and interrupts;
-- `grading.json`: assertion evidence;
-- `artifacts/`: the subject workspace output.
-
-`compare` exits nonzero on a gated regression, making it suitable for CI.
-
-## Regrade and annotate
-
-```bash
-lux grade <run>
-lux grade <run> --case cases/changed-grader.ts
-lux annotate <run> \
-  --out evals/annotations.json \
-  --rubric evals/rubrics/case.json \
-  --label pass
-```
-
-Regrading preserves subject evidence and changes only grading. Human annotations
-support grader-alignment experiments. `annotate` can record overall label or
-score, repeatable assertion outcomes, split/tag selection, feedback, annotator,
-reviewer, and human/synthetic provenance. Keep assertion IDs aligned with the
-rubric and record reviewer context where available.
-
-## Experiment lifecycle
-
-```bash
-lux experiment <campaign> --plan
-lux experiment <campaign>
-lux experiment <campaign> --resume <experiment-dir>
-lux optimize <campaign> --plan
-lux optimize <campaign>
-lux optimize <campaign> --resume <experiment-dir>
-lux view <experiment-dir>
-lux report <experiment-dir> --out .lux-runs/reports/decision.md
-lux report <experiment-dir> --json
-lux apply <experiment-dir>
-```
-
-`experiment` compares authored variants. `optimize` searches proposed changes.
-Both create durable, resumable, integrity-checked records. `report` re-verifies
-the ledger and promotion decision. `apply` is deliberately separate and
-hash-guards the source it changes.
-
-Use the exact resume, report, and apply commands printed by Lux. Do not edit
-`.lux-runs/.experiments/` manually.
-
-## Failure triage
-
-- Discovery/config error: run `doctor`, verify paths and exported definitions.
-- Driver unavailable: verify the selected CLI is installed and the model is
-  explicitly configured where required.
-- Agent exit/error: inspect `events.jsonl` and the driver section of `run.json`.
-- Assertion failure: inspect the assertion evidence and final artifact; decide
-  whether the subject or measurement is wrong.
-- Tool evidence mismatch: compare paired call/result IDs and inspect unresolved,
-  unmatched, malformed, and failed tool outcomes; never infer success from a
-  call event alone.
-- Judge disagreement: narrow the rubric, regrade stored evidence, and collect
-  human annotations before changing the subject.
-- Interrupted experiment: resume the printed experiment directory; do not start
-  a nominally identical campaign and combine results yourself.