Turn every new podcast episode into a week's worth of content — show notes, quote graphics, social clips, and a ready-to-send email newsletter — automatically

On first run the agent asks the user:

• (1) the source of new episodes — any combination of podcast RSS (the user pastes the show's RSS URL and the agent watches for new items), YouTube (the agent watches a channel or playlist for new uploads — YouTube subtitle download is only available on a local self-hosted IronClaw because YouTube blocks cloud / VPS IPs, so this option is gated behind a self-host setup check), and manual upload (the user drops an MP3 / WAV file into the agent through any connected chat channel).
• (2) How to get the transcript — the agent tries the cheapest path first: if the RSS enclosure includes a transcript URL, the agent pulls that directly; if the YouTube upload has subtitles and the user is on a local self-hosted IronClaw, the agent pulls those; otherwise the agent runs Whisper transcription on the audio.
• (3) Which outputs to generate (the user can pick any combination — show notes is on by default; quote graphics prompts, social clips, and email newsletter are optional add-ons). Show notes are a 5–10 bullet summary with chapter markers and timestamps, quote graphics are 5–10 pull-quotes each paired with an image-generation prompt (for Midjourney, DALL-E, or a designer), social clips are 3–5 short posts formatted per platform (Twitter / X thread, LinkedIn post, Instagram caption), and the email newsletter is a ready-to-send HTML or plain-text draft with a one-line subject, preview text, intro, key takeaways, and a CTA.
• (4) The default delivery channel — the user is asked to confirm which channel the agent should drop new episode outputs into, defaulting to the channel the user is most active in. From there the user can also opt into a Notion database (one row per episode with all outputs as fields), a Google Drive folder (one folder per episode, with show_notes.md, quotes.md, clips.md, newsletter.md as separate files), or just the in-channel summary.
• (5) Whether to push a one-line notification to the channel the moment an episode finishes processing, or stay silent and let the user pull.

These preferences are stored. On every new episode the agent pulls the audio or transcript from the configured source, runs Whisper if needed, and produces the enabled outputs in parallel. Each output is written into the chosen destination and a one-line notification is sent if the user opted in.

The user can at any time ask the agent to regenerate a specific output ("give me a punchier version of the show notes", "draft 3 more quote graphics", "rewrite the newsletter in a more casual tone"), change the RSS feed, swap delivery channel, or add a new output type.