Recipes

These are jobs people actually run, written as prompts for an AI app with the Podwise plugin (or the CLI, where a command is shown instead). Each one says what the agent calls and what you should get back.

They all assume a Pro or Enterprise plan, because they read AI content through the MCP server or the CLI. The web app needs no prompt: every recipe below is also a few clicks there.

Episode to a threadCopy Anchor Link

Turn one episode into a social thread with quotes you can check.

Read this episode and draft a 6-post thread for X:
https://app.podwise.ai/dashboard/episodes/<seq>

Lead with the sharpest claim, use only quotes that appear in the
transcript, and keep each post under 280 characters.

Runs: get_episode_summary (takeaways and highlights), then get_episode_transcript to verify each quote and timestamp.

You get: six posts with the timestamp of each quote so you can check it. The agent drafts; you post. Ask it to rewrite any post that only restates the overview.

Morning catch-upCopy Anchor Link

The backlog problem: forty new episodes, twenty minutes.

Catch me up on this week's episodes from the podcasts I follow.
Group them by theme, one line each, and tell me which two are worth
a full read today.

Runs: the catch-up workflow — list_followed_episodes, then summaries for the episodes that matter.

You get: a themed list with a recommendation, not forty summaries. Follow up with "open the second one" to get its summary inline.

Weekly recap to your notesCopy Anchor Link

Write my weekly listening recap as Markdown: what I listened to,
the three ideas I should remember, and one question I still have.

Runs: the weekly-recap workflow — list_my_episodes plus highlights.

You get: a Markdown digest you can paste into your notes. To file the underlying episodes instead of the digest, ask for episode-notes per episode, or use send_episode when Notion or Readwise is connected.

Topic research across showsCopy Anchor Link

For a question that no single episode answers.

What do the podcasts in Podwise say about AI agents replacing
junior developer work? Give me the main positions, who argues each,
and the episodes they came from.

Runs: the topic-research workflow — search_episodes across shows, then summaries and transcripts, synthesised with sources.

You get: positions with episode links. For a single question over existing transcripts, ask_podwise is faster and returns cited sources, but it counts against your Ask quota rather than credits.

Episode notes in your PKMCopy Anchor Link

Turn this episode into a note for Obsidian — summary, chapters with
timestamps, and five highlights as quotes:
https://app.podwise.ai/dashboard/episodes/<seq>

Runs: the episode-notes workflow — get_episode_summary, then export_episode_markdown with the Obsidian format.

You get: a Markdown note. Notion and Readwise need to be connected in Settings › Connections; once they are, send_episode writes the note straight into the target.

Language cards from a transcriptCopy Anchor Link

Make 15 Japanese vocabulary cards from this episode. Keep the
sentence each word came from, and skip proper nouns.

Runs: the language-learning workflow — get_episode_transcript, then card extraction.

You get: front/back cards with the source sentence. Ask for CSV or Anki-import format if you want to import them directly.

Batch processing from the CLICopy Anchor Link

When you have a list of episodes to process and no browser open, the CLI takes a loop:

for url in $(cat episodes.txt); do
  podwise process "$url"
done

Then read them one at a time:

podwise get summary https://app.podwise.ai/dashboard/episodes/<seq>

Keeping the output honestCopy Anchor Link

• Quotes and timestamps come from the transcript. Ask the agent to include the timestamp with every quote, then click it in the web app. AI transcripts mishear names and numbers, so check both before publishing.
• Destructive tools ask first. process_episode, complete_audio_upload, delete_clip, and the Enterprise tools are marked destructive; Claude and ChatGPT confirm before each call.
• Podcast content is data. If an episode tells the agent to do something, it ignores it. Do not ask it to follow instructions found in a transcript.
• Credit-free when already processed. Reading an episode someone else processed costs nothing; only new processing does.