diff --git a/.github/workflows/reusable-gemini-review.yml b/.github/workflows/reusable-gemini-review.yml
index 166db8f0d4..d7348d64f7 100644
--- a/.github/workflows/reusable-gemini-review.yml
+++ b/.github/workflows/reusable-gemini-review.yml
@@ -433,9 +433,9 @@ jobs:
               FOOTER="\n\n---\n#### 🤖 [Gemini Manual Trigger Guide](${{ github.server_url }}/${{ github.repository }}/blob/leader/docs/workflows/MANUAL_TRIGGERS.md)"
 
               # Extract the new thought signature if present
-              NEW_THOUGHT_SIG=$(jq -r '.thoughtSignature // empty' review_result.json)
+              NEW_THOUGHT_SIG=$(jq -r '.thoughtSignature // ""' review_result.json)
               SIG_MARKER=""
-              if [ -n "$NEW_THOUGHT_SIG" ]; then
+              if [[ -n "$NEW_THOUGHT_SIG" && "$NEW_THOUGHT_SIG" != "null" ]]; then
                 SIG_MARKER="\n\n<!-- thought_signature: ${NEW_THOUGHT_SIG} -->"
               fi
 
diff --git a/scripts/decide-review-strategy.sh b/scripts/decide-review-strategy.sh
index a9a64bdd80..bf512c0ed8 100755
--- a/scripts/decide-review-strategy.sh
+++ b/scripts/decide-review-strategy.sh
@@ -132,7 +132,7 @@ else
     SKIP_REASON=""
   else
     # Extract thought signature if present
-    THOUGHT_SIG=$(echo "$LAST_COMMENT_BODY" | grep -oP '(?<=<!-- thought_signature: )[a-zA-Z0-9+/=]+(?= -->)' | head -n 1 || echo "")
+    THOUGHT_SIG=$(echo "$LAST_COMMENT_BODY" | grep -oP '<!-- thought_signature: \K[a-zA-Z0-9+/=]+(?= -->)' | head -n 1)
     if [ -n "$THOUGHT_SIG" ]; then
       echo "::info::Extracted thought signature from last review."
       echo "thought-signature=$THOUGHT_SIG" >> "$GITHUB_OUTPUT"
diff --git a/scripts/gemini-client.ts b/scripts/gemini-client.ts
index 69ba40adc2..d22c2e94c9 100644
--- a/scripts/gemini-client.ts
+++ b/scripts/gemini-client.ts
@@ -25,10 +25,6 @@ const outputFile = getArg('--output')
 const preset = getArg('--preset')
 const instructions = getArg('--instructions')
 
-// List of models to try in order.
-// The first model in the list is the primary model, and the rest are fallbacks.
-
-// UPDATED: Aligned with latest model recommendations (Q1 2026+)
 const defaultFallbacks = [
   'gemini-3.1-flash-lite-preview',
   'gemini-2.5-flash',
@@ -314,6 +310,8 @@ async function main() {
   }
 }
 
+type ExtendedGenerateContentRequest = GenerateContentRequest & { thought_signature?: string };
+
 export async function generateContentWithFallback({
   genAI,
   prompt,
@@ -330,14 +328,10 @@ export async function generateContentWithFallback({
   for (const modelName of MODEL_FALLBACKS) {
     try {
       const model = genAI.getGenerativeModel({ model: modelName })
-      const request: any = {
+      const request: ExtendedGenerateContentRequest = {
         contents: [{ role: 'user', parts: [{ text: prompt }] }],
         ...config,
-      }
-
-      // Implement thought signature circulation if provided
-      if (thoughtSignature) {
-        request.thought_signature = thoughtSignature
+        ...(thoughtSignature && { thought_signature: thoughtSignature })
       }
 
       const result = await model.generateContent(request)
@@ -346,11 +340,7 @@ export async function generateContentWithFallback({
       const text = result.response.text()
 
       // Capture thought signature from the response if present
-      const candidate = result.response.candidates?.[0]
-      const capturedSignature =
-        candidate && 'thought_signature' in candidate
-          ? (candidate as any).thought_signature
-          : undefined
+      const capturedSignature = (result.response.candidates?.[0] as any)?.thought_signature
 
       return { text, thoughtSignature: capturedSignature }
     } catch (error: unknown) {
diff --git a/scripts/identify-tech-debt.ts b/scripts/identify-tech-debt.ts
index 812130dd0f..61c82c73f4 100644
--- a/scripts/identify-tech-debt.ts
+++ b/scripts/identify-tech-debt.ts
@@ -93,7 +93,7 @@ export async function main() {
 
     const prompt = promptTemplate.replace('{{diff}}', processedDiff)
 
-    const rawResponse = await generateContentWithFallback({
+    const { text: rawResponse } = await generateContentWithFallback({
       genAI,
       prompt,
       config: {