Add type-specific outcome evaluators for create_issue, add_comment, and add_labels

.github/workflows/requirements.txt

@@ -1,5 +1,4 @@
 "mempalace==3.2.0"
-\\\n
 markitdown-mcp
 numpy
 scikit-learn

actions/setup/js/add_labels.cjs

@@ -177,6 +177,20 @@ const main = createCountGatedHandler({
         };
       }
 
+      /** @type {string[]} */
+      let labelsBefore = [];
+      try {
+        const { data: existingItem } = await githubClient.rest.issues.get({
+          owner: repoParts.owner,
+          repo: repoParts.repo,
+          issue_number: itemNumber,
+        });
+        const existingLabels = Array.isArray(existingItem?.labels) ? existingItem.labels : [];
+        labelsBefore = existingLabels.map(label => (typeof label === "string" ? label : label?.name)).filter(name => typeof name === "string" && name.trim() !== "");
+      } catch (error) {
+        core.info(`Unable to capture labels-before snapshot for ${contextType} #${itemNumber}: ${getErrorMessage(error)}`);
+      }
+
       try {
         await withRetry(
           () =>
@@ -195,6 +209,7 @@ const main = createCountGatedHandler({
           success: true,
           number: itemNumber,
           labelsAdded: uniqueLabels,
+          labelsBefore,
           contextType,
         };
       } catch (error) {