diff --git a/pyproject.toml b/pyproject.toml
index 6e3d41a..57735f3 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -97,10 +97,14 @@ DEP002 = [
     "pytest-cov",
     "ruff",
 ]
-# DEP001: scripts/import_chrome_session.py imports browser-cookie3 behind a
-# guarded `try/except ImportError` — it is a deliberately optional helper
-# dependency, installed on demand, not a declared runtime requirement.
+# DEP001 + DEP003: scripts/import_chrome_session.py + import_safari_session.py
+# import browser-cookie3 behind a guarded `try/except ImportError` — it is a
+# deliberately optional helper dependency, installed on demand, not a declared
+# runtime requirement. Deptry sees it as both undeclared (DEP001) and, once
+# pulled in via another package's extras on some machines, as transitive
+# (DEP003). Both are intentional for this opt-in helper.
 DEP001 = ["browser_cookie3"]
+DEP003 = ["browser_cookie3"]
 
 [tool.coverage.run]
 source = ["src/xbrain"]
diff --git a/src/xbrain/archive.py b/src/xbrain/archive.py
index 5e27f61..7e5091c 100644
--- a/src/xbrain/archive.py
+++ b/src/xbrain/archive.py
@@ -50,17 +50,13 @@ def _parse_js_array(raw: str, tweets_file: str) -> list:
         raise ValueError(f"{tweets_file}: malformed JSON in archive tweets file: {exc}") from exc
 
 
-def _archive_tweet_to_item(entry: dict[str, Any], author: Author) -> Item | None:
-    tweet = entry.get("tweet")
-    if not isinstance(tweet, dict):
-        logger.warning("archive entry missing 'tweet' object, skipping")
-        return None
-    rest_id = tweet.get("id_str")
-    if not rest_id:
-        logger.warning("archive tweet missing 'id_str', skipping")
-        return None
-    rest_id = str(rest_id)
-    links = [
+def _extract_archive_links(tweet: dict[str, Any]) -> list[Link]:
+    """Parse `entities.urls` from an archive tweet into Link objects.
+
+    Each URL entity in the archive carries an `expanded_url`; we derive the
+    domain from it and drop any entry that lacks an expanded URL.
+    """
+    return [
         Link(
             url=url_entity["expanded_url"],
             domain=urlparse(url_entity["expanded_url"]).netloc,
@@ -68,10 +64,19 @@ def _archive_tweet_to_item(entry: dict[str, Any], author: Author) -> Item | None
         for url_entity in tweet.get("entities", {}).get("urls", [])
         if url_entity.get("expanded_url")
     ]
+
+
+def _extract_archive_media(tweet: dict[str, Any]) -> list[MediaEntry]:
+    """Parse media from `extended_entities.media`, falling back to `entities.media`.
+
+    Normalises `type` to `"video"` (videos and animated GIFs) or `"photo"`, and
+    picks the canonical URL (`media_url_https`, falling back to `expanded_url`).
+    Entries missing both URLs are dropped.
+    """
     media_entries = tweet.get("extended_entities", {}).get("media") or tweet.get(
         "entities", {}
     ).get("media", [])
-    media: list[MediaEntry] = [
+    return [
         Media(
             type="video" if media_entity.get("type") in ("video", "animated_gif") else "photo",
             url=media_entity.get("media_url_https") or media_entity["expanded_url"],
@@ -79,6 +84,19 @@ def _archive_tweet_to_item(entry: dict[str, Any], author: Author) -> Item | None
         for media_entity in media_entries
         if media_entity.get("media_url_https") or media_entity.get("expanded_url")
     ]
+
+
+def _archive_tweet_to_item(entry: dict[str, Any], author: Author) -> Item | None:
+    """Map one archive entry into an Item, returning None for malformed rows."""
+    tweet = entry.get("tweet")
+    if not isinstance(tweet, dict):
+        logger.warning("archive entry missing 'tweet' object, skipping")
+        return None
+    rest_id = tweet.get("id_str")
+    if not rest_id:
+        logger.warning("archive tweet missing 'id_str', skipping")
+        return None
+    rest_id = str(rest_id)
     return Item(
         id=rest_id,
         source="own_tweet",
@@ -87,7 +105,7 @@ def _archive_tweet_to_item(entry: dict[str, Any], author: Author) -> Item | None
         text=tweet.get("full_text", ""),
         created_at=_parse_x_date(tweet.get("created_at")),
         captured_at=datetime.now(timezone.utc),
-        media=media,
-        links=links,
+        media=_extract_archive_media(tweet),
+        links=_extract_archive_links(tweet),
         quoted_id=None,
     )
diff --git a/src/xbrain/generate.py b/src/xbrain/generate.py
index 83d6574..1e242ec 100644
--- a/src/xbrain/generate.py
+++ b/src/xbrain/generate.py
@@ -353,16 +353,27 @@ def _frontmatter(item: Item) -> str:
     return "\n".join(lines)
 
 
-def _render_index(items: list[Item], strings: Strings) -> str:
-    bookmarks = sum(1 for i in items if i.source == "bookmark")
-    own = sum(1 for i in items if i.source == "own_tweet")
-    noted = sum(1 for i in items if _has_note(i))
-    enriched = sum(1 for i in items if i.enriched)
+def _count_topic_frequency(items: list[Item]) -> dict[str, int]:
+    """Tally how often each topic appears across the enriched items.
+
+    Items without enrichment contribute nothing. The result maps topic slug
+    to the number of enriched items that include it.
+    """
     topic_freq: dict[str, int] = {}
     for item in items:
         if item.enriched:
             for topic in item.enriched.topics:
                 topic_freq[topic] = topic_freq.get(topic, 0) + 1
+    return topic_freq
+
+
+def _render_index(items: list[Item], strings: Strings) -> str:
+    """Render the top-level index note: corpus stats and the topic list."""
+    bookmarks = sum(1 for i in items if i.source == "bookmark")
+    own = sum(1 for i in items if i.source == "own_tweet")
+    noted = sum(1 for i in items if _has_note(i))
+    enriched = sum(1 for i in items if i.enriched)
+    topic_freq = _count_topic_frequency(items)
     lines = [
         "# XBrain",
         "",
diff --git a/src/xbrain/validate.py b/src/xbrain/validate.py
index 44e63c3..a5afe05 100644
--- a/src/xbrain/validate.py
+++ b/src/xbrain/validate.py
@@ -14,46 +14,75 @@
 _ALLOWED_KEYS = {"summary", "primary_topic", "topics"}
 
 
-def validate_judgment(judgment: dict, vocab_slugs: Iterable[str]) -> list[str]:
-    """Return a list of human-readable errors; an empty list means valid."""
-    rules = load_guardrails().get("enrichment", {})
-    vocab = set(vocab_slugs)
-    errors: list[str] = []
-
+def _validate_judgment_keys(judgment: dict) -> list[str]:
+    """Reject any key outside the allowed enrichment schema."""
     extra = set(judgment) - _ALLOWED_KEYS
     if extra:
-        errors.append(f"unexpected keys (LLM must emit only judgment): {sorted(extra)}")
+        return [f"unexpected keys (LLM must emit only judgment): {sorted(extra)}"]
+    return []
 
+
+def _validate_summary(judgment: dict, rules: dict) -> list[str]:
+    """Require a non-empty summary when guardrails demand it."""
     summary = judgment.get("summary")
     if rules.get("summary_required", True) and not (summary and str(summary).strip()):
-        errors.append("summary is missing or empty")
+        return ["summary is missing or empty"]
+    return []
+
 
+def _validate_topics_list(judgment: dict, rules: dict, vocab: set[str]) -> list[str]:
+    """Validate the topics list itself: count bounds, duplicates, vocabulary membership.
+
+    Returns an empty list when `topics` is not a list — the caller (`validate_judgment`)
+    is responsible for emitting the type error and aborting further topic-related checks.
+    """
     topics = judgment.get("topics")
     if not isinstance(topics, list):
-        errors.append("topics must be a list")
-        return errors
-
+        return []
+    errors: list[str] = []
     lo, hi = rules.get("topics_min", 1), rules.get("topics_max", 4)
     if not (lo <= len(topics) <= hi):
         errors.append(f"topics has {len(topics)} entries, must be {lo}-{hi}")
-
     if len(set(topics)) != len(topics):
         errors.append("topics has duplicate entries")
-
     if rules.get("topics_must_be_in_vocab", True):
         for slug in topics:
             if slug not in vocab:
                 errors.append(f"topic '{slug}' is not in the vocabulary")
+    return errors
+
 
+def _validate_primary_topic(
+    judgment: dict, topics: list, rules: dict, vocab: set[str]
+) -> list[str]:
+    """Validate `primary_topic`: presence, vocabulary membership, and inclusion in topics."""
     primary = judgment.get("primary_topic")
     if not primary:
-        errors.append("primary_topic is missing")
-    else:
-        if rules.get("topics_must_be_in_vocab", True) and primary not in vocab:
-            errors.append(f"primary_topic '{primary}' is not in the vocabulary")
-        if rules.get("primary_topic_must_be_in_topics", True) and primary not in topics:
-            errors.append(f"primary_topic '{primary}' is not inside topics")
+        return ["primary_topic is missing"]
+    errors: list[str] = []
+    if rules.get("topics_must_be_in_vocab", True) and primary not in vocab:
+        errors.append(f"primary_topic '{primary}' is not in the vocabulary")
+    if rules.get("primary_topic_must_be_in_topics", True) and primary not in topics:
+        errors.append(f"primary_topic '{primary}' is not inside topics")
+    return errors
+
+
+def validate_judgment(judgment: dict, vocab_slugs: Iterable[str]) -> list[str]:
+    """Return a list of human-readable errors; an empty list means valid."""
+    rules = load_guardrails().get("enrichment", {})
+    vocab = set(vocab_slugs)
+
+    errors: list[str] = []
+    errors += _validate_judgment_keys(judgment)
+    errors += _validate_summary(judgment, rules)
+
+    topics = judgment.get("topics")
+    if not isinstance(topics, list):
+        errors.append("topics must be a list")
+        return errors
 
+    errors += _validate_topics_list(judgment, rules, vocab)
+    errors += _validate_primary_topic(judgment, topics, rules, vocab)
     return errors