diff --git a/pyproject.toml b/pyproject.toml
index 48fc992..92f5851 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -1,6 +1,6 @@
 [project]
 name = "pixelml-av"
-version = "0.1.0"
+version = "0.2.0"
 description = "Index once, search many — agentic video memory CLI by Pixel ML"
 readme = "README.md"
 requires-python = ">=3.11"
@@ -37,5 +37,21 @@ build-backend = "hatchling.build"
 [tool.hatch.build.targets.wheel]
 packages = ["src/av"]
 
+[tool.hatch.build.targets.wheel.force-include]
+"prompts/dense_caption.md" = "av/prompts/dense_caption.md"
+
+[tool.hatch.build.targets.sdist]
+include = [
+    "/src",
+    "/tests",
+    "/cookbook",
+    "/prompts",
+    "/skills",
+    "/README.md",
+    "/LICENSE",
+    "/pyproject.toml",
+    "/uv.lock",
+]
+
 [tool.pytest.ini_options]
 testpaths = ["tests"]
diff --git a/skills/agentic-video-memory/SKILL.md b/skills/agentic-video-memory/SKILL.md
index cb61ed9..72f21ac 100644
--- a/skills/agentic-video-memory/SKILL.md
+++ b/skills/agentic-video-memory/SKILL.md
@@ -1,6 +1,6 @@
 ---
 name: agentic-video-memory
-version: 0.1.0
+version: 0.2.0
 description: Video memory for AI agents. Ingest, search, and ask questions over video content.
 homepage: https://agentic.video
 ---
@@ -382,7 +382,7 @@ Print version info.
 **Output (stdout, JSON):**
 
 ```json
-{"version": "0.1.0", "package": "pixelml-av"}
+{"version": "0.2.0", "package": "pixelml-av"}
 ```
 
 ## Agent Integration Patterns
diff --git a/src/av/__init__.py b/src/av/__init__.py
index 492034f..603f820 100644
--- a/src/av/__init__.py
+++ b/src/av/__init__.py
@@ -1,3 +1,3 @@
 """av — Agentic Video CLI by Pixel ML. Index once, search many."""
 
-__version__ = "0.1.0"
+__version__ = "0.2.0"
diff --git a/src/av/cli/ask.py b/src/av/cli/ask.py
index eedd678..ff21b07 100644
--- a/src/av/cli/ask.py
+++ b/src/av/cli/ask.py
@@ -26,7 +26,14 @@ def ask_cmd(
             help="Skip configured Jev relevance and scene refinement",
         ),
     ) -> None:
-        """Ask a question about indexed video content (RAG)."""
+        """Ask a question about indexed video content (RAG).
+
+        When configured, Jev filters source relevance, bounds scene context, and
+        checks answer support. Optional stronger selected-frame inspection runs
+        only with explicit AV_STRONG_VISION_API_BASE_URL and
+        AV_STRONG_VISION_MODEL settings; AV_STRONG_VISION_API_KEY is used when
+        the endpoint requires authentication.
+        """
         config = get_config(db_path=Path(db) if db else None)
         repo = Repository(config.db_path)
 
diff --git a/src/av/pipeline/ingest.py b/src/av/pipeline/ingest.py
index 2e83dc9..55ad171 100644
--- a/src/av/pipeline/ingest.py
+++ b/src/av/pipeline/ingest.py
@@ -27,6 +27,15 @@
 from av.utils.principles import load_principles
 
 
+def _dense_caption_template_path() -> Path:
+    current = Path(__file__).resolve()
+    candidates = (
+        current.parents[1] / "prompts" / "dense_caption.md",
+        current.parents[3] / "prompts" / "dense_caption.md",
+    )
+    return next((candidate for candidate in candidates if candidate.is_file()), candidates[-1])
+
+
 def _chunk_seconds_for_audio(audio_path: Path, target_bytes: int = MAX_AUDIO_CHUNK_BYTES - (1 * 1024 * 1024)) -> int:
     size = audio_path.stat().st_size
     if size <= 0:
@@ -387,7 +396,7 @@ def ingest_video(
                     )
 
                 principles = load_principles(principles_path)
-                template_path = Path(__file__).resolve().parents[3] / "prompts" / "dense_caption.md"
+                template_path = _dense_caption_template_path()
                 prompt = render_dense_prompt(template_path, principles)
 
                 if frames:
diff --git a/uv.lock b/uv.lock
index 112cebd..d625e68 100644
--- a/uv.lock
+++ b/uv.lock
@@ -352,7 +352,7 @@ wheels = [
 
 [[package]]
 name = "pixelml-av"
-version = "0.1.0"
+version = "0.2.0"
 source = { editable = "." }
 dependencies = [
     { name = "openai" },