Daily AI morning brief plus a 90-second explainer video, rendered on the bot's cloud computer

(Translated from the author's Chinese.) Please set up an "AI morning brief + explainer video" pipeline for me: every day, compile the previous day's AI news into a text brief, then turn it into a ~90-second Chinese explainer video. First install the environment, do a trial run with yesterday's content, and hand it to me for review.
[1. Install skills (on your own computer)]
Clone the three public repos below into ~/.claude/skills/ and read each SKILL.md and the files it references in full:
- (link) (video pipeline: React + GSAP for visuals, Playwright frame-by-frame render, FFmpeg compositing)
- (link) (layout: Swiss style, paper-white background, safety orange #FF6B35)
- (link) (de-AI the narration)
Install missing dependencies per the skills' environment checks (Node, Playwright + Chromium, FFmpeg, Python, etc.). Also install edge-tts for Chinese voice-over; start with zh-CN-YunxiNeural at +20% speed; swap in a more stable TTS if you have one.
[2. Collect news (date Y = yesterday in my time zone)]
- Use one full day in my time zone, converted to UTC for queries; don't use UTC midnight directly.
- Sources: my X bookmarks, my own posts on day Y (top 3 by likes + 2×replies + 2×bookmarks, with substantive views from the replies), and the other sources I list below.
- My X account: @<your username>
- Other sources (optional): <e.g. a newsletter's sender address, a web page>
- If a source can't be read, say "couldn't read" honestly, separately from "no update".
[3. Text brief]
- Group related content into 5–7 storylines, each with a one-sentence headline, one or two key points and key numbers.
- Attach the original post link to each (…/<username>/status/<id>); if you added background, also attach an official source.
- Check numbers like prices, benchmarks and funding against official pages. Mark rumors "unconfirmed / community rumor". Never invent numbers, quotes, links or usernames; fill in reply authors by post id.
[4. Explainer video]
- Follow the Product Video skill's process exactly: write DIRECTION.md first, then build shots, then render. Fonts, colors and layout follow the social card skill.
- Structure: intro (AI Morning Brief + date + item count) → one screen per story (headline, 1–2 orange key numbers, source account) → an X radar screen → outro. 60–90 seconds total, landscape 1920×1080 by default.
- Conversational Chinese narration, run through Humanizer-zh first. Synthesize each sentence separately and measure its length; align shots, subtitles and motion to the actual voice-over durations. Burn in subtitles and duck the music under the voice.
- Only include facts already in the brief; add nothing new.
- After export, run the skill's delivery check and review sampled frames screen by screen: no overflowing, overlapping or cut-off text, no missing Chinese glyphs, consistent color across the film. Fix problems before delivering.
[5. Delivery]
Send me: the full text brief, the final mp4, the narration script, 3–4 key frames, plus one sentence on which steps are still unstable. Don't send any messages or posts on my behalf.
Once I'm happy with the trial run, set up a scheduled task that runs automatically at 9am every morning.