fix: prevent contentless follow-up issues from verifier

.github/scripts/verifier_issue_formatter.js

@@ -4,6 +4,51 @@ const {
   parseScopeTasksAcceptanceSections,
 } = require('./issue_scope_parser.js');
 
+/**
+ * Check if content is a placeholder (missing section marker).
+ * These placeholders are generated by the bot when PR/issue lacks structured content.
+ *
+ * @param {string} text - Text to check
+ * @returns {boolean} True if text is a placeholder
+ */
+function isPlaceholderContent(text) {
+  const normalized = String(text || '').toLowerCase().trim();
+  // Match common placeholder patterns
+  return /^[-*]?\s*\[[\sx]\]\s*(scope|tasks?|acceptance\s*criteria?)\s+(section\s+)?missing\s+from\s+source\s+issue\.?$/.test(normalized) ||
+    /section\s+missing\s+from\s+source\s+issue/.test(normalized) ||
+    /^n\s*\/?\s*a$/.test(normalized) ||
+    normalized === '';
+}
+
+/**
+ * Check if an item looks like a markdown section header that was incorrectly
+ * captured as a list item (e.g., "## Related" or "### Notes").
+ *
+ * @param {string} text - Text to check
+ * @returns {boolean} True if text looks like a section header
+ */
+function looksLikeSectionHeader(text) {
+  const trimmed = String(text || '').trim();
+  // Match markdown headers at start
+  return /^#{1,6}\s+\w/.test(trimmed);
+}
+
+/**
+ * Check if an item looks like a PR/Issue reference link rather than actual criteria.
+ * Matches patterns like:
+ * - "- PR #123 - Title"
+ * - "- Issue #456 - Description"  
+ * - "PR #789 - Some fix"
+ *
+ * @param {string} text - Text to check
+ * @returns {boolean} True if text looks like a reference link
+ */
+function looksLikeReferenceLink(text) {
+  const trimmed = String(text || '').trim();
+  // Match "- PR #N" or "- Issue #N" or just "PR #N" / "Issue #N" at start
+  return /^[-–•]?\s*(PR|Issue|Pull\s+Request)\s*#\d+/i.test(trimmed);
+}
+
 /**
  * Simple similarity score between two strings (0-1).
  * Uses Jaccard similarity on word sets for fuzzy matching.
@@ -171,6 +216,9 @@ function parseVerifierFindings(verifierOutput) {
 
 /**
  * Extract unchecked items from a checkbox section.
+ * Filters out generic noise such as placeholder content, section headers, and
+ * reference links so it can be reused for any checkbox-based content
+ * (e.g., criteria, tasks, or other checklist items).
  *
  * @param {string} content - Section content with checkboxes
  * @returns {string[]} Array of unchecked item texts
@@ -184,14 +232,22 @@ function extractUncheckedItems(content) {
     const match = line.match(/^\s*[-*]\s+\[\s\]\s+(.+)$/);
     if (match) {
       const text = match[1].trim();
-      // Skip markdown section headers that were incorrectly captured as criteria
-      if (text.startsWith('#') || text.startsWith('##')) {
+
+      // Skip placeholder content (e.g., "Tasks section missing from source issue.")
+      if (isPlaceholderContent(text)) {
         continue;
       }
-      // Skip items that look like markdown links to sections (e.g., "- PR #123")
-      if (/^[-–]\s*(PR|Issue)\s*#\d+/i.test(text)) {
+
+      // Skip markdown section headers (e.g., "## Related")
+      if (looksLikeSectionHeader(text)) {
         continue;
       }
+
+      // Skip PR/Issue reference links (e.g., "- PR #123 - Title")
+      if (looksLikeReferenceLink(text)) {
+        continue;
+      }
+
       items.push(text);
     }
   }
@@ -452,12 +508,21 @@ function formatFollowUpIssue({
     ? `[Follow-up] Unmet criteria from PR #${prNumber}`
     : '[Follow-up] Verifier failure - unmet acceptance criteria';
 
+  // Determine if this issue has substantive content worth creating
+  // Skip if we have no real tasks/criteria (just defaults) and no verifier gaps
+  const hasSubstantiveContent =
+    (newTasks.length > 0 && !newTasks.every(t => isPlaceholderContent(t))) ||
+    (refinedUnmetCriteria.length > 0 && !refinedUnmetCriteria.every(c => isPlaceholderContent(c))) ||
+    findings.gaps.length > 0 ||
+    findings.unmetCriteria.length > 0;
+
   return {
     title,
     body: sections.join('\n\n'),
     findings,
     unmetCriteria: refinedUnmetCriteria,
     newTasks,
+    hasSubstantiveContent,
   };
 }
 
@@ -516,10 +581,18 @@ function formatSimpleFollowUpIssue({
     ? `Verifier failure for PR #${prNumber}`
     : 'Verifier failure on merged commit';
 
+  // Simple format always has substantive content (verifier output)
+  // since we only use it when we have actual verifier output to display
+  const hasSubstantiveContent = 
+    findings.gaps.length > 0 || 
+    findings.unmetCriteria.length > 0 ||
+    (verifierOutput && verifierOutput.trim().length > 0);
+
   return {
     title,
     body: lines.join('\n'),
     findings,
+    hasSubstantiveContent,
   };
 }
 
@@ -531,4 +604,8 @@ module.exports = {
   extractUncheckedItems,
   extractCheckedItems,
   buildChecklist,
+  // Helper functions exported for testing
+  isPlaceholderContent,
+  looksLikeSectionHeader,
+  looksLikeReferenceLink,
 };

.github/workflows/agents-verifier.yml

@@ -353,13 +353,23 @@ jobs:
               });
             }
 
+            // Skip issue creation if there's no substantive content
+            // (e.g., all sections are placeholder text or empty)
+            if (!result.hasSubstantiveContent) {
+              core.info('Skipping issue creation: no substantive content to act on (all placeholder sections)');
+              core.setOutput('issue_number', '');
+              core.setOutput('skipped', 'true');
+              return;
+            }
+
             const { data: issue } = await github.rest.issues.create({
               ...context.repo,
               title: result.title,
               body: result.body,
               labels: ['agent:codex'],
             });
             core.setOutput('issue_number', issue?.number ? String(issue.number) : '');
+            core.setOutput('skipped', 'false');
             core.info(`Created follow-up issue #${issue.number}: ${result.title}`);
 
       - name: Collect verifier metrics

.github/workflows/reusable-agents-verifier.yml

@@ -320,13 +320,23 @@ jobs:
               });
             }
 
+            // Skip issue creation if there's no substantive content
+            // (e.g., all sections are placeholder text or empty)
+            if (!result.hasSubstantiveContent) {
+              core.info('Skipping issue creation: no substantive content to act on (all placeholder sections)');
+              core.setOutput('issue_number', '');
+              core.setOutput('skipped', 'true');
+              return;
+            }
+
             const { data: issue } = await github.rest.issues.create({
               ...context.repo,
               title: result.title,
               body: result.body,
               labels: ['agent:codex'],
             });
             core.setOutput('issue_number', issue?.number ? String(issue.number) : '');
+            core.setOutput('skipped', 'false');
             core.info(`Created follow-up issue #${issue.number}: ${result.title}`);
 
       - name: Collect verifier metrics

docs/workflows/SystemEvaluation.md

@@ -71,8 +71,8 @@ This document evaluates the current GitHub Actions workflow architecture against
 **Goal**: Identify gaps in the current workflow suite and opportunities for speed optimization.
 
 *   **Performance**:
-    *   **Sequential Testing**: `reusable-10-ci-python.yml` runs `pytest` without parallelism. `pytest-xdist` is missing from `pyproject.toml`.
-        *   **Impact**: Tests run one by one, slowing down feedback loops as the test suite grows.
+    *   **Parallel Testing**: ✅ `reusable-10-ci-python.yml` now runs `pytest -n auto --dist=loadgroup` when `pytest-xdist` is detected. The workflow automatically installs `pytest-xdist` and enables parallelism.
+        *   **Status**: Implemented. Tests run in parallel across available CPU cores.
     *   **Artifact Overload**: The CI workflow uploads a massive amount of artifacts (coverage XML/JSON/HTML, metrics, history, classification, delta, trend) for *every* run.
         *   **Impact**: Increases storage costs and workflow runtime (upload/download time).
     *   **Fail-Slow Strategy**: The CI workflow uses `continue-on-error: true` for all checks (lint, format, typecheck, test) to gather a full report.
@@ -88,7 +88,7 @@ This document evaluates the current GitHub Actions workflow architecture against
     *   **Gap**: It does not check for "Stale PRs" or "Dependency Freshness" (beyond what Dependabot might do silently).
 
 *   **Recommendations**:
-    *   **Enable Parallel Testing**: Add `pytest-xdist` to `pyproject.toml` and update `reusable-10-ci-python.yml` to use `pytest -n auto`.
+    *   ~~**Enable Parallel Testing**: Add `pytest-xdist` to `pyproject.toml` and update `reusable-10-ci-python.yml` to use `pytest -n auto`.~~ ✅ Done
     *   **Optimize Artifacts**: Only upload full coverage reports on `main` or when explicitly requested. Use summary comments for PRs.
     *   **Add Security Workflow**: Create `security.yml` running `bandit` and `safety`.
     *   **Create Release Workflow**: Add `release.yml` for automated publishing.