diff --git a/crypto/stark/src/tests/small_trace_tests.rs b/crypto/stark/src/tests/small_trace_tests.rs
index 96e04858d..ea8d3bc4a 100644
--- a/crypto/stark/src/tests/small_trace_tests.rs
+++ b/crypto/stark/src/tests/small_trace_tests.rs
@@ -116,6 +116,34 @@ fn test_verify_rejects_truncated_composition_poly_parts_ood() {
     );
 }
 
+/// A malformed proof whose `deep_poly_openings` Vec is shorter than the FRI
+/// query count. `reconstruct_deep_composition_poly_evaluations_for_all_queries`
+/// indexes `deep_poly_openings[i]` for every query index, and this Vec's length
+/// is not otherwise bound (the `query_list.len()` guard checks a different
+/// field), so a truncated `deep_poly_openings` must make the verifier return
+/// `false` instead of panicking with an out-of-bounds index in release builds.
+#[test_log::test]
+fn test_verify_rejects_truncated_deep_poly_openings() {
+    let (air, mut proof) = make_valid_simple_proof();
+
+    assert!(
+        proof.deep_poly_openings.len() >= 2,
+        "test precondition: a valid proof has one deep-poly opening per FRI query",
+    );
+    // Drop the last opening so the Vec is shorter than `fri_number_of_queries`;
+    // the query loop would then index past the end.
+    proof.deep_poly_openings.pop();
+
+    assert!(
+        !Verifier::verify(
+            &proof,
+            &air,
+            &mut DefaultTranscript::<GoldilocksField>::new(&[])
+        ),
+        "Verifier must reject when deep_poly_openings is shorter than the query count"
+    );
+}
+
 /// A malformed proof whose deep-poly opening `evaluations` slice has the
 /// wrong number of columns. The runtime width-mismatch guard added in this
 /// PR must cause the verifier to return `false` instead of indexing past
diff --git a/crypto/stark/src/verifier.rs b/crypto/stark/src/verifier.rs
index 616732e22..a9dc8f381 100644
--- a/crypto/stark/src/verifier.rs
+++ b/crypto/stark/src/verifier.rs
@@ -539,6 +539,18 @@ pub trait IsStarkVerifier<
         proof: &StarkProof<Field, FieldExtension, PI>,
     ) -> Option<DeepPolynomialEvaluations<FieldExtension>> {
         let num_queries = challenges.iotas.len();
+
+        // `deep_poly_openings` comes straight from the untrusted proof and its
+        // length is not otherwise pinned (the `query_list.len()` guard checks a
+        // different field). The loop below indexes `deep_poly_openings[i]` for
+        // every `i` in `0..num_queries`, so a truncated Vec would panic the
+        // verifier with an out-of-bounds index on a malicious proof. Reject
+        // instead. (Extra entries are harmless — they are never indexed —
+        // matching the `<` convention of the `query_list` guard.)
+        if proof.deep_poly_openings.len() < num_queries {
+            return None;
+        }
+
         let mut deep_poly_evaluations = Vec::with_capacity(num_queries);
         let mut deep_poly_evaluations_sym = Vec::with_capacity(num_queries);