diff --git a/CLAUDE.md b/CLAUDE.md
index 797c1109..b9fb103b 100644
--- a/CLAUDE.md
+++ b/CLAUDE.md
@@ -75,7 +75,13 @@ Read that file to access the full content.
 ./scripts/build-index.sh
 ```
 
-This calls Claude CLI for each episode to generate keywords. The script is idempotent - it skips episodes already present in keyword files, so it can be run multiple times safely.
+This calls Claude CLI (using `claude-sonnet-4-20250514`) for each episode to generate keywords. The script is idempotent - it skips episodes already present in keyword files, so it can be run multiple times safely.
+
+**Environment variables** (optional):
+
+- `EPISODES_DIR` - Source directory (default: `episodes`)
+- `OUTPUT_DIR` - Index output directory (default: `index`)
+- `TEMP_DIR` - Temp file location (default: system temp)
 
 ## Adding Publication Dates
 
@@ -83,8 +89,10 @@ All episodes should include `publish_date` in ISO 8601 format (YYYY-MM-DD). To f
 
 1. Use the `video_id` from the transcript's frontmatter
 2. Call the YouTube Data API v3:
-   ```
+
+   ```text
    https://www.googleapis.com/youtube/v3/videos?part=snippet&id={video_id}&key={API_KEY}
    ```
+
 3. Extract `snippet.publishedAt` from the response
 4. Add `publish_date: YYYY-MM-DD` to the frontmatter after `video_id`