diff --git a/.github/workflows/release.yml b/.github/workflows/release.yml
index aa3c6a8e3..3d9b71a82 100644
--- a/.github/workflows/release.yml
+++ b/.github/workflows/release.yml
@@ -244,21 +244,27 @@ jobs:
         if: ${{ inputs.use_ai_notes }}
         shell: bash
         env:
-          MODELS_TOKEN: ${{ secrets.MODELS_TOKEN }}
+          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
         run: |
           set -euo pipefail
 
           python <<'PY'
           from pathlib import Path
           import json
-          import os
 
           prompt = Path(".github/release-artifacts/release-prompt.txt").read_text(encoding="utf-8")
           payload = {
-              "model": "gpt-5.4",
+              "model": "openai/gpt-5.4",
               "messages": [
-                  {"role": "user", "content": prompt}
-              ],
+                  {
+                      "role": "system",
+                      "content": "You write concise, accurate release notes in markdown."
+                  },
+                  {
+                      "role": "user",
+                      "content": prompt
+                  }
+              ]
           }
           Path(".github/release-artifacts/openai-payload.json").write_text(
               json.dumps(payload, ensure_ascii=False),
@@ -266,11 +272,23 @@ jobs:
           )
           PY
 
-          curl -fsSL https://models.inference.ai.azure.com/chat/completions \
-            -H "Content-Type: application/json" \
-            -H "Authorization: Bearer $MODELS_TOKEN" \
-            -d @.github/release-artifacts/openai-payload.json \
-            > .github/release-artifacts/openai-response.json
+          status_code=$(
+            curl -sS -L \
+              -o .github/release-artifacts/openai-response.json \
+              -w "%{http_code}" \
+              -X POST \
+              -H "Accept: application/vnd.github+json" \
+              -H "Authorization: Bearer $GITHUB_TOKEN" \
+              -H "X-GitHub-Api-Version: 2026-03-10" \
+              -H "Content-Type: application/json" \
+              https://models.github.ai/orgs/dataelement/inference/chat/completions \
+              -d @.github/release-artifacts/openai-payload.json
+          )
+
+          echo "HTTP status: $status_code"
+          cat .github/release-artifacts/openai-response.json
+
+          test "$status_code" -lt 400
 
           python <<'PY'
           from pathlib import Path
diff --git a/backend/app/services/trigger_runtime/executions.py b/backend/app/services/trigger_runtime/executions.py
index c25882565..bda34f1c0 100644
--- a/backend/app/services/trigger_runtime/executions.py
+++ b/backend/app/services/trigger_runtime/executions.py
@@ -84,8 +84,10 @@ async def claim_pending_trigger_executions(
             claimed_pairs.append((execution, trigger))
         await db.commit()
         for execution, trigger in claimed_pairs:
-            db.expunge(execution)
-            db.expunge(trigger)
+            if execution in db:
+                db.expunge(execution)
+            if trigger in db:
+                db.expunge(trigger)
     return claimed_pairs
 
 
diff --git a/backend/app/services/trigger_runtime/invoker.py b/backend/app/services/trigger_runtime/invoker.py
index 56f29a49d..9c585c8ff 100644
--- a/backend/app/services/trigger_runtime/invoker.py
+++ b/backend/app/services/trigger_runtime/invoker.py
@@ -103,15 +103,21 @@ async def invoke_agent_for_triggers(agent_id: uuid.UUID, triggers: list[AgentTri
             result = await db.execute(select(Agent).where(Agent.id == agent_id))
             agent = result.scalar_one_or_none()
             if not agent or agent.is_expired:
+                if execution_ids:
+                    await mark_trigger_executions_failed(execution_ids, "Agent not found or is expired")
                 return
 
             if not agent.primary_model_id:
                 logger.warning(f"Agent {agent.name} has no LLM model, skipping trigger invocation")
+                if execution_ids:
+                    await mark_trigger_executions_failed(execution_ids, "Agent has no LLM model configured")
                 return
             result = await db.execute(select(LLMModel).where(LLMModel.id == agent.primary_model_id))
             model = result.scalar_one_or_none()
             if not model or not model.enabled:
                 logger.warning(f"Agent {agent.name}'s model is unavailable, skipping trigger invocation")
+                if execution_ids:
+                    await mark_trigger_executions_failed(execution_ids, "Agent primary model is unavailable or disabled")
                 return
 
             context_parts = []
diff --git a/frontend/src/components/MarkdownRenderer.tsx b/frontend/src/components/MarkdownRenderer.tsx
index 48f44e6fe..4c38df536 100644
--- a/frontend/src/components/MarkdownRenderer.tsx
+++ b/frontend/src/components/MarkdownRenderer.tsx
@@ -106,10 +106,10 @@ function renderInline(text: string): string {
         .replace(/\*\*\*(.*?)\*\*\*/g, '<strong><em>$1</em></strong>')
         // Bold
         .replace(/\*\*(.*?)\*\*/g, '<strong>$1</strong>')
-        .replace(/__(.*?)__/g, '<strong>$1</strong>')
+        .replace(/(?<!\w)__(?!\s)(.+?)(?<!\s)__(?!\w)/g, '<strong>$1</strong>')
         // Italic
         .replace(/\*(.*?)\*/g, '<em>$1</em>')
-        .replace(/_(.*?)_/g, '<em>$1</em>')
+        .replace(/(?<!\w)_(?!\s)(.+?)(?<!\s)_(?!\w)/g, '<em>$1</em>')
         // Strikethrough
         .replace(/~~(.*?)~~/g, '<del>$1</del>');