@pome-sh/cli 0.23.32 → 0.23.33

package/dist/build-info.json

@@ -1,6 +1,6 @@
 {
   "package": "pome-sh",
-  "version": "0.23.32",
-  "git_sha": "07c526747a7654a4aea3b1d8f417da3c9d653238",
-  "build_time": "2026-08-13T01:00:51.116Z"
+  "version": "0.23.33",
+  "git_sha": "224f20c230aa9a79b6e074916240940da903dcaf",
+  "build_time": "2026-08-13T01:09:10.392Z"
 }

package/dist/src/cli/main.js

@@ -3064,7 +3064,22 @@ async function runChecksCommand(twinArg, opts) {
 // src/task/insertCriterion.ts
 var HEADING_RE = /^##\s+Success Criteria\s*$/;
 var NEXT_HEADING_RE = /^##\s+/;
-var CRITERION_RE = /^[-*]\s+\[(?:code|model)(?::[a-z][a-z0-9_-]*)?(?:\s+always-scored)?\]\s+.+$/;
+var CRITERION_RE = /^[-*]\s+\[(code|model)(?::([a-z][a-z0-9_-]*))?(\s+always-scored)?\]\s+(.+)$/;
+function identityKey(criterion) {
+  return `${criterion.kind}\0${criterion.twin ?? ""}\0${criterion.text}`;
+}
+function readCriterionLine(line) {
+  const match = line.trim().match(CRITERION_RE);
+  if (!match) return void 0;
+  return { kind: match[1], tag: match[2], text: match[4].trim() };
+}
+function primaryTwin(source) {
+  try {
+    return readConfigTwins(source)[0];
+  } catch {
+    return void 0;
+  }
+}
 var MissingCriteriaSectionError = class extends Error {
   constructor(path) {
     super(
@@ -3074,11 +3089,15 @@ var MissingCriteriaSectionError = class extends Error {
   }
 };
 var DuplicateCriterionError = class extends Error {
-  constructor(path, line) {
+  constructor(path, line, existing = line) {
+    const differs = existing.trim() !== line.trim();
     super(
       `${path} already has this criterion:
-  ${line}
-Adding it again would score it twice and inflate the task's denominator.`
+  ${existing.trim()}
+` + (differs ? `which is the same check as the one being added:
+  ${line.trim()}
+\u2014 a marker annotation or a twin tag does not make it a second check.
+` : "") + `Adding it again would score it twice and inflate the task's denominator.`
     );
     this.name = "DuplicateCriterionError";
   }
@@ -3094,13 +3113,33 @@ function insertCriterion(source, line, path = "this task file") {
       break;
     }
   }
+  const primary = primaryTwin(source);
+  const declared = /* @__PURE__ */ new Map();
+  for (const criterion of readCodeCriteria(source)) {
+    declared.set(
+      identityKey({ kind: "code", twin: criterion.twin, text: criterion.text }),
+      `- ${criterion.marker} ${criterion.text}`
+    );
+  }
   let insertAt = -1;
   for (let i = start + 1; i < end; i += 1) {
-    const existing = lines[i].trim();
-    if (!CRITERION_RE.test(existing)) continue;
-    if (existing === line.trim()) throw new DuplicateCriterionError(path, line);
+    const parsed = readCriterionLine(lines[i]);
+    if (!parsed) continue;
+    if (parsed.kind === "model") {
+      declared.set(
+        identityKey({ kind: "model", twin: parsed.tag ?? primary, text: parsed.text }),
+        lines[i].trim()
+      );
+    }
     insertAt = i + 1;
   }
+  const incoming = readCriterionLine(line);
+  if (incoming) {
+    const stored = declared.get(
+      identityKey({ kind: incoming.kind, twin: incoming.tag ?? primary, text: incoming.text })
+    );
+    if (stored !== void 0) throw new DuplicateCriterionError(path, line, stored);
+  }
   if (insertAt === -1) {
     insertAt = lines[start + 1]?.trim() === "" ? start + 2 : start + 1;
   }
@@ -4541,7 +4580,7 @@ var DEFAULT_AGENT_COMMAND = `node ${DEFAULT_AGENT_FILE}`;
 var MANIFEST_SCHEMA_URL = "https://pome.sh/schemas/v1/pome.json";
 var MAX_UNREADABLE_PATHS_SHOWN = 5;
 function readPackageVersion() {
-  if ("0.23.32".length > 0) return "0.23.32";
+  if ("0.23.33".length > 0) return "0.23.33";
   try {
     const here = dirname(fileURLToPath(import.meta.url));
     const candidates = [

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@pome-sh/cli",
-  "version": "0.23.32",
+  "version": "0.23.33",
   "description": "Digital-twin testing for AI agents \u2014 run tasks against resettable local or hosted twins and record tool-call traces for evaluation on pome.sh.",
   "keywords": [
     "ai",