diff --git a/src/skillspector/llm_analyzer_base.py b/src/skillspector/llm_analyzer_base.py
index c5ab9dce..4835c97b 100644
--- a/src/skillspector/llm_analyzer_base.py
+++ b/src/skillspector/llm_analyzer_base.py
@@ -46,6 +46,38 @@
 CHARS_PER_TOKEN = 4
 CHUNK_OVERLAP_LINES = 50
 
+_MEDIA_FILE_EXTENSIONS = (
+    ".png",
+    ".jpg",
+    ".jpeg",
+    ".gif",
+    ".bmp",
+    ".ico",
+    ".webp",
+    ".avif",
+    ".heic",
+    ".heif",
+    ".tif",
+    ".tiff",
+    ".mp3",
+    ".aac",
+    ".flac",
+    ".m4a",
+    ".ogg",
+    ".opus",
+    ".wav",
+    ".mp4",
+    ".m4v",
+    ".avi",
+    ".mov",
+    ".webm",
+    ".mkv",
+    ".mpeg",
+    ".mpg",
+    ".ogv",
+    ".3gp",
+)
+
 
 # ---------------------------------------------------------------------------
 # Default structured-output schemas (discovery mode)
@@ -303,6 +335,9 @@ def get_batches(
 
         batches: list[Batch] = []
         for path in file_paths:
+            if path.lower().endswith(_MEDIA_FILE_EXTENSIONS):
+                logger.info("Skipping media file from LLM analysis: %s", path)
+                continue
             content = file_cache.get(path) or "No content available for this file."
             file_findings = findings_by_file.get(path, [])
 
diff --git a/tests/nodes/test_llm_analyzer_base.py b/tests/nodes/test_llm_analyzer_base.py
index e344e654..6b928c64 100644
--- a/tests/nodes/test_llm_analyzer_base.py
+++ b/tests/nodes/test_llm_analyzer_base.py
@@ -203,6 +203,44 @@ def test_zero_padding(self) -> None:
         assert "L11: line10" in result
 
 
+# ---------------------------------------------------------------------------
+# LLMAnalyzerBase.get_batches
+# ---------------------------------------------------------------------------
+
+
+class TestLLMAnalyzerBaseGetBatches:
+    MODEL = "nvidia/openai/gpt-oss-120b"
+
+    @pytest.mark.parametrize(
+        "path",
+        [
+            "assets/demo.gif",
+            "assets/screenshot.PNG",
+            "assets/tutorial.mp4",
+            "assets/voice.mp3",
+            "assets/photo.webp",
+            "assets/movie.mkv",
+        ],
+    )
+    @patch(MOCK_PATCH_TARGET, _mock_get_chat_model)
+    def test_media_files_are_skipped(self, path: str) -> None:
+        analyzer = LLMAnalyzerBase(base_prompt="test", model=self.MODEL)
+
+        assert analyzer.get_batches([path], {path: "decoded media data"}) == []
+
+    @patch(MOCK_PATCH_TARGET, _mock_get_chat_model)
+    def test_text_files_and_svg_are_preserved(self) -> None:
+        analyzer = LLMAnalyzerBase(base_prompt="test", model=self.MODEL)
+        file_cache = {
+            "src/main.py": "print('hello')\n",
+            "assets/icon.svg": '<svg><script>alert("x")</script></svg>',
+        }
+
+        batches = analyzer.get_batches(list(file_cache), file_cache)
+
+        assert {batch.file_path for batch in batches} == set(file_cache)
+
+
 # ---------------------------------------------------------------------------
 # LLMAnalyzerBase.build_prompt (default implementation)
 # ---------------------------------------------------------------------------