50 Claude Prompts That Produce a Podcast

You are a senior interview producer who builds run sheets that keep an hour of tape usable. <context> I record an interview this week and I want a run sheet instead of a list of questions I forget to follow. </context> <inputs> - Guest name and role: [DETAILS] - What makes them worth an hour: [SPECIFIC WORK] - Episode topic: [TOPIC] - Target length: [MINUTES] - What the listener should be able to do after: [OUTCOME] - Points I must hit: [LIST] - CTA or sponsor slots: [WHERE] </inputs> <task> Build a minute by minute run sheet: the moment to listen for as a cold open, the host intro read, the opening question, four or five topic blocks each with a time budget and the one point it must land, the transition line out of each block, the question to save for late, and the closing sequence with the ask. Then mark where to cut if I am running long. </task> <constraints> - Each block needs exactly one must land point stated in plain words. - Time budgets must add up to the target length with five minutes of slack. - No question that their own website already answers. </constraints> <format> Return the run sheet as a two column artifact with time and content, then list the five questions to drop first if time runs short. </format>

You are an interviewer known for getting specifics out of media trained executives. <context> My guest gives polished answers and I need questions that push past their standard talking points. </context> <inputs> - Guest and role: [DETAILS] - Their claim to expertise: [WHAT THEY DID] - Episode topic: [TOPIC] - What my audience wants to steal from them: [OUTCOME] - Their known talking points: [LIST OR PASTE A PAST INTERVIEW] - Areas that are off limits: [LIST] </inputs> <task> Write 18 questions arranged in four rungs: warm up questions that get them talking in specifics, story questions that force a concrete scene, mechanism questions about how the thing actually worked including numbers and sequence, and pressure questions that test the claim. For each question state the answer I am fishing for and the follow up to use if the answer comes back generic. </task> <constraints> - No yes or no questions and no questions answerable from their homepage. - Include one genuinely hard question that is fair rather than hostile. - Mechanism questions must ask for a number, a date, or a name. </constraints> <format> Return a table as an artifact with question, what I am fishing for, and the follow up, then the order I should actually ask them in. </format>

You are a scriptwriter for solo audio who writes for the ear and not for the page. <context> I am recording a solo episode teaching one thing and I do not want it to sound like someone reading a blog post. </context> <inputs> - The one thing I am teaching: [SKILL OR IDEA] - Who is listening: [AUDIENCE] - Why I am credible on this: [ONE LINE] - The framework or steps: [3 TO 4 STEPS] - A real example I can tell: [SHORT DESCRIPTION] - Target length: [MINUTES] </inputs> <task> Write a full solo script: a hook, the promise, one line on why me, the framework in three or four steps each anchored to a concrete example, the mistake most people make, a short live walk through, a recap, and the call to action. Write it to be spoken, with an estimated read time per section. </task> <constraints> - Sentences under 20 words, spoken register, contractions allowed. - Never read a bullet list aloud. Turn every list into spoken sequence. - One concrete example per abstraction with no exceptions. </constraints> <format> Return the script as an artifact with pause markers and emphasis marks, then the section by section timing estimate. </format>

You are an audio editor who chooses cold opens that stop the skip. <context> The first 30 seconds decide whether anyone hears the rest of this episode and I keep defaulting to a bland welcome. </context> <inputs> - Episode topic and guest: [DETAILS] - Transcript or notes: [PASTE] - The single most surprising thing in the episode: [DETAIL] - Audience: [WHO] - Episode promise: [ONE LINE] - House intro I use today: [PASTE OR NONE] </inputs> <task> Write eight cold open options, each using a different mechanism: a surprising number, a contrarian claim, a guest quote pulled forward, starting mid story, a listener question, the stakes, a before and after, and a myth to break. Each option is 20 to 35 seconds with the exact words and the line that comes immediately after it. </task> <constraints> - Every option must make sense to someone who has never heard the show. - If a quote works as the cold open, use the guest's real words from my transcript. - No option may promise something the episode does not deliver. </constraints> <format> Return the eight options as an artifact labelled by mechanism, then your pick with the reasoning in two lines. </format>

You are a moderator who runs multi guest audio without letting one voice eat the hour. <context> I have three or four guests on one episode and unmoderated panels turn into mush. </context> <inputs> - Panelists and their positions: [NAME, ROLE, VIEW] - Topic: [TOPIC] - Where I expect them to disagree: [DETAILS] - Target length: [MINUTES] - What the listener should leave with: [OUTCOME] - Topics to steer away from: [LIST] </inputs> <task> Build a moderation plan: the introduction order and the framing for each guest, a disagreement map showing who diverges from whom and on what, the questions engineered to surface that tension usefully, an air time balance strategy with the exact wording I can use to interrupt or redirect, and a closing sequence that forces a position instead of ending on vague agreement. </task> <constraints> - Give me literal interrupt lines I can say out loud without sounding rude. - Every question must be pointed at a named panelist and say why them. - The ending must produce a disagreement summary, not a round of thanks. </constraints> <format> Return the run sheet as an artifact plus a table of question, target panelist, and the reason it goes to them. </format>

You are a research producer who briefs hosts so they never sound like the least informed person in the room. <context> I am recording on a topic I know only partly and I need to be credible without faking expertise. </context> <inputs> - Episode topic: [TOPIC] - Guest if any: [NAME AND ROLE] - Audience and their level: [WHO, HOW EXPERT] - What I already know: [SUMMARY] - Angle I want to take: [ANGLE] - Recording date: [DATE] </inputs> <task> Write a pre record brief: the 10 things I must know, the current state of the debate with the main camps and who holds them, the numbers that get cited in this space and where they originally come from, five questions nobody answers well, the jargon I should define for a general listener, and three concrete examples I can reference. Then flag everything that needs verification before I say it on tape. </task> <constraints> - Label every claim you are not confident about and never invent a statistic. - Say explicitly where your knowledge is likely out of date. - Separate settled facts from contested positions. </constraints> <format> Return the brief as an artifact, then a short verify before recording list with the exact thing to check for each item. </format>

You are an interview coach who trains hosts to hear a non answer and push in real time. <context> My guests give smooth answers and I only notice the empty ones while editing, which is too late. </context> <inputs> - Guest and topic: [DETAILS] - Their likely talking points: [LIST OR PASTE A PAST INTERVIEW] - What my audience needs to be concrete: [OUTCOME] - How confrontational I am willing to be: [SCALE] - Recording format: [REMOTE / IN PERSON] - Episode length: [MINUTES] </inputs> <task> For each likely talking point, write three probes: one that demands a specific number, date, or name, one that asks for the counterexample or the time it did not work, and one that asks for the mechanism behind the claim. Then build a one page cheat sheet of universal probes I can keep on screen while recording. </task> <constraints> - Probes must be short enough to say in one breath while listening. - Each probe stays polite while making a vague answer impossible to repeat. - The cheat sheet must fit on one screen in large text. </constraints> <format> Return the probe table as an artifact, then the cheat sheet formatted in about ten lines I can read at a glance. </format>

You are a debate coach who steel mans positions rather than knocking down straw men. <context> I hold a strong position on this topic and I want to argue the other side honestly on air. </context> <inputs> - My position: [CLAIM] - Why I believe it: [EVIDENCE] - Audience: [WHO] - Who disagrees and why: [CAMPS] - Episode length: [MINUTES] - How much I am willing to concede: [SCOPE] </inputs> <task> Build the strongest honest case against my position: the five best counterarguments ranked by force, the kind of evidence behind each, the two places my position is genuinely weak, the steel manned version of the opposing view stated as its holders would state it, and a segment structure for arguing it on air. Then tell me which counterargument I currently cannot answer. </task> <constraints> - Steel man every counterargument. No caricatures. - Name the type of evidence behind each point and its limits. - Write the exact concession lines I should say out loud. </constraints> <format> Return the brief as an artifact with the ranked counterarguments, then the segment plan and the concession lines. </format>

You are a producer who turns customer stories into episodes that do not sound like sponsored content. <context> I want to run a case study episode with a customer or operator without it becoming an hour long advert. </context> <inputs> - Guest and company: [DETAILS] - The before state: [SITUATION] - What they changed: [DECISION] - Results with numbers: [METRICS OR UNKNOWN] - What went wrong along the way: [OBSTACLES] - Who in my audience this helps: [WHO] </inputs> <task> Structure a case study episode in beats: the before scene with a specific moment, the trigger that forced a decision, the decision and the tradeoffs rejected, what broke during execution, the numbers, what they would do differently, and how a listener applies it on Monday. For each beat write the questions that get that beat on tape, and give me a compliance safe way to handle metrics they cannot share. </task> <constraints> - Every beat needs a scene or a number, never a summary. - The what broke beat is mandatory and must be at least five minutes. - Provide the exact line to say if the guest cannot disclose figures. </constraints> <format> Return the run sheet as an artifact with beat, purpose, and questions, then the metrics fallback lines. </format>

You are an editor who turns a pile of listener questions into a paced episode. <context> I have a stack of listener questions and dumping them in order would make a boring episode. </context> <inputs> - Questions I received: [PASTE] - Recurring themes I noticed: [LIST] - Episode length: [MINUTES] - Positions the show already holds: [LIST] - Audience: [WHO] - Where questions come from: [EMAIL, DMS, FORM] </inputs> <task> Group the questions into themes, select the six to eight worth airing and say why the rest are out, order them for pacing with quick wins early and the meatiest in the middle and the most memorable last, rewrite each in one line so it makes sense to listeners who did not ask, draft the spine of each answer with one concrete example, and write the call to action for more questions. </task> <constraints> - Merge near duplicate questions instead of answering them twice. - Each answer spine must contain one example or one number. - Give a time budget per question that fits the episode length. </constraints> <format> Return the run sheet as an artifact with columns for question, rewritten version, answer spine, and time budget. </format>

You are a booker who lands hard to reach guests with short, specific emails. <context> I want to invite someone who does not know me and my last few invitations went unanswered. </context> <inputs> - Guest name and role: [DETAILS] - The specific thing they did that prompted this: [DETAIL] - Show premise and audience: [WHO LISTENS] - Episode angle only they can cover: [ANGLE] - Time ask and format: [E.G. 45 MINUTES REMOTE] - Proof the show is real: [DOWNLOADS, PAST GUESTS, BRAND] </inputs> <task> Write a cold invitation under 150 words: a subject line, an opening sentence proving I know their specific work, the angle framed as good for them, the concrete ask with length and format and two date options, one line of proof the show is real, and a single easy reply instruction. Then give two alternate subject lines and a 40 word direct message version. </task> <constraints> - No flattery paragraph and no asking to pick their brain. - The angle must be something only this person can talk about. - One ask only. No secondary requests hidden at the end. </constraints> <format> Return the email as an artifact, then the alternates, then a short note on what to personalise for each new guest. </format>

You are a talent researcher who finds guests who have done the work rather than guests who talk about it. <context> I am short on guests for the next quarter and my obvious list is already booked or over exposed. </context> <inputs> - Season theme or topic: [TOPIC] - Audience: [WHO LISTENS] - Guest criteria: [SENIORITY, SECTOR, REGION] - Guests I already had on: [LIST] - Types of guest to avoid: [LIST] - Effort I can spend on outreach: [HOURS PER WEEK] </inputs> <task> Propose 25 candidates in three tiers: reach, realistic, and hidden gem. For each give the reason they fit, the angle only they can cover, a guess at how reachable they are, and a possible warm path in. Flag anyone who is likely over exposed on the podcast circuit and say so. </task> <constraints> - Never repeat a guest I listed as already booked. - Prefer operators and practitioners over professional speakers. - Say clearly when you are unsure a person is current or real so I verify. </constraints> <format> Return the candidates as a table artifact grouped by tier, then the five I should contact first and in what order. </format>

You are a producer who writes the document that makes a busy expert say yes without a call. <context> Good guests want to know what they are walking into before they give me an hour. </context> <inputs> - Show name and premise: [DETAILS] - Who listens: [ROLES, SENIORITY, SIZE] - What guests get: [AUDIENCE, CLIPS, LINKS] - Recording setup: [PLATFORM, LENGTH, PREP, GEAR] - Past guests I can name: [LIST OR NONE] - Our editorial rules: [WHAT WE WILL AND WILL NOT DO] </inputs> <task> Write a one page guest document covering: what the show is, who actually listens with real detail, what a guest gets out of it, exactly what recording involves step by step, what we will not do such as ambush questions or publishing without notice, past guests if any, and the next step in one sentence. Then compress it into a five line version I can paste into an email. </task> <constraints> - Everything must fit on one page with no filler paragraphs. - State audience quality honestly rather than inflating numbers. - The what we will not do section is mandatory and specific. </constraints> <format> Return the one pager as a clean document artifact, then the five line email version. </format>

You are a research producer who prepares hosts so a guest hears a question nobody has asked them. <context> I record with this guest soon and I do not want to ask what every other host already asked. </context> <inputs> - Guest name, role, company: [DETAILS] - Material I pasted: [BIO, POSTS, PAST INTERVIEWS, TALKS] - Episode topic: [TOPIC] - What I want my audience to learn: [OUTCOME] - Sensitive areas I know about: [LIST OR NONE] - Time I have to prepare: [MINUTES] </inputs> <task> Build a dossier from the material I pasted: a three line bio, the arc of their career, their signature ideas and the exact phrases they use, claims they have repeated across interviews, the stories they have already told to death, likely sensitivities, and six opening questions grounded in specific things they have said or built. </task> <constraints> - Use only the material I pasted plus clearly labelled inference. - Never invent a quote, a date, or a job title. - Flag anything I should confirm with the guest before recording. </constraints> <format> Return the dossier as an artifact, then a short do not ask list of the stories and questions they have already answered many times. </format>

You are a production coordinator who sends the email that makes guests arrive prepared and calm. <context> Guests keep showing up with bad audio and no story ready, and it costs me in the edit. </context> <inputs> - Guest name: [NAME] - Confirmed time and their timezone: [DETAILS] - Recording platform and link: [DETAILS] - Areas we will cover: [3 TO 4 AREAS] - My actual first question: [QUESTION] - What they should have ready: [NUMBERS, STORY, EXAMPLE] - What happens after recording: [TIMELINE] </inputs> <task> Write the pre record email: the confirmed time in their timezone, the link, the length, the areas we will cover with my literal first question, the gear and setup requests, quick room and microphone tips, what to have ready, how to handle a fumble mid answer, what happens after recording, and who to contact on the day. Then write a shorter 24 hour reminder version. </task> <constraints> - Under 300 words with the essentials in the first half. - Friendly and practical. Nothing that reads like a legal form. - Include a copyable checklist block of five items maximum. </constraints> <format> Return the email as an artifact with the checklist block, then the 24 hour reminder version. </format>

You are a booker who converts silence into bookings without becoming annoying. <context> Half my invitations get no reply and I need a follow up sequence that does not just bump the thread. </context> <inputs> - Guest and role: [DETAILS] - Original invitation I sent: [PASTE] - Days since I sent it: [NUMBER] - New value I can add: [E.G. RECENT EPISODE, AUDIENCE DATA, THEIR NEWS] - Lower friction options I can offer: [SHORTER SLOT, ASYNC, WRITTEN] - Deadline if any: [DATE OR NONE] </inputs> <task> Write a three touch follow up: a day four nudge that adds something new, a day twelve message offering a lower friction version of the ask, and a day twenty five graceful close that leaves the door open for later. Then give me routing rules for the common replies including interested but busy, wrong topic, send questions first, and hard no, plus the scheduling message to send on a yes. </task> <constraints> - Each touch must add new information. No message that only bumps. - No guilt, no false deadlines, no reply expected language. - The final message must make a future yes easy without another pitch. </constraints> <format> Return the three emails as an artifact, then a reply handling table mapping each likely response to the next move. </format>

You are an editorial gatekeeper who screens inbound guest pitches before they waste an episode. <context> Agencies pitch me guests every week and most would produce a forgettable hour. </context> <inputs> - Show premise and audience: [DETAILS] - What makes a great guest for me: [TRAITS] - Recent pitches I regretted: [WHAT WENT WRONG] - Non negotiables: [E.G. NO PRODUCT PITCHES] - How many episodes I can fill from inbound: [NUMBER PER MONTH] - Who screens pitches: [ME / ASSISTANT] </inputs> <task> Build a scorecard with eight weighted criteria covering specific expertise, evidence of real operating experience, willingness to be concrete, audience fit, promotional reach, freshness on the circuit, and reputational risk. Define what a low, medium, and high score looks like for each, set pass, hold, and decline thresholds, and write five screening questions to send before booking anyone. </task> <constraints> - Every criterion must be scorable from a short email exchange or a call. - Weights must total 100 and reflect what actually predicts a good episode. - The screening questions must be answerable in under five minutes. </constraints> <format> Return the scorecard as a table artifact, then the screening email and the threshold rules. </format>

You are a producer who protects editorial control while staying on good terms with PR teams. <context> I get more inbound guest pitches than I can handle and I answer each one from scratch. </context> <inputs> - Show premise and audience: [DETAILS] - Typical pitch I receive: [PASTE ONE] - My editorial conditions: [E.G. NO SCRIPT APPROVAL] - What I would say yes to: [CRITERIA] - Tone I want to keep: [E.G. WARM BUT FIRM] - Who sends the replies: [ME / ASSISTANT] </inputs> <task> Write five reply templates, each under 120 words: a clean decline, a decline with a redirect to a better fit, a request for a screening call, a counter proposing a different angle with the same person, and an acceptance that states conditions such as no script approval, no product pitch, and figures on the record. Then give a one line rule for when to use each. </task> <constraints> - Kind but final where the answer is no. No false hope. - Conditions must be stated without sounding adversarial. - Each template must be usable by an assistant with light editing. </constraints> <format> Return the five templates as an artifact, then the selection rules as a short table. </format>

You are a producer who turns a finished recording into a guest who actually promotes the episode. <context> My guests say they will share the episode and then usually do not. </context> <inputs> - Guest name: [NAME] - One specific thing that was great in the interview: [DETAIL] - Publish date: [DATE] - What I will send them: [CLIPS, QUOTE CARDS, LINKS] - Where they should share: [PLATFORMS] - Tagging and link details: [HANDLES, URL] </inputs> <task> Write the post recording email: one specific piece of praise that proves I listened, what happens next with dates, the assets I will send, the promotional ask framed as easy including a pre written post they can edit, the tagging and link details, and an invitation to flag anything they want cut. Then write the publish day follow up that delivers the assets. </task> <constraints> - The promotional ask must include the draft copy, not a request to write their own. - Under 200 words for the first email. - Offer the cut option genuinely, with a deadline. </constraints> <format> Return both emails as an artifact, with the draft post for the guest included in full. </format>

You are a studio producer who runs the five minutes before record that decide how the hour goes. <context> I start recordings awkwardly and I forget to get consent and levels sorted before the good stuff happens. </context> <inputs> - Show name: [NAME] - Guest and topic: [DETAILS] - Platform and setup: [DETAILS] - Usage rights I need: [AUDIO, VIDEO, CLIPS, TRANSCRIPT] - How I handle mistakes and restarts: [POLICY] - My opening question: [QUESTION] </inputs> <task> Write the pre roll the host runs in the five minutes before recording: the greeting, the level check phrases, the reassurance that mistakes get edited out, the consent and usage line to capture on tape, the check on anything they do not want covered, how we will start, the agreed signal for a restart, and the last minute question to loosen them up. Then write my own host checklist for the same five minutes. </task> <constraints> - The consent line must be short enough to say naturally on tape. - Level check phrases must contain plosives and sibilants, not just hello. - The host checklist must be under ten items and in running order. </constraints> <format> Return the pre roll script as an artifact, then the host checklist formatted so I can print it. </format>

You are a podcast editor who writes show notes people actually read instead of skim past. <context> I have a finished transcript and I need publish ready show notes without inventing anything. </context> <inputs> - Transcript: [PASTE] - Guest and role: [DETAILS] - Episode number and title: [DETAILS] - Links I have: [GUEST, RESOURCES] - House style notes: [TONE, LENGTH, FORMAT] - Call to action: [WHAT I WANT] </inputs> <task> Produce publish ready show notes: a 40 word hook, a three sentence summary, five to seven takeaways written as actual claims rather than topic labels, a timestamped chapter list, notable quotes with timestamps, every resource mentioned with its correct name, the guest links, and the call to action. Then flag anything you could not source in the transcript. </task> <constraints> - Use only what is in the transcript and never invent a link or a figure. - Takeaways must state the claim, for example that hiring before product market fit doubled their burn, not the word hiring. - Keep the guest's own wording where it is stronger than yours. </constraints> <format> Return the notes as a markdown artifact ready to paste, then a short list of gaps for me to fill in. </format>

You are a title writer who balances what people search against what makes them click. <context> My episode titles are guest names and topic labels, and nobody outside my subscribers ever clicks them. </context> <inputs> - Episode topic and guest: [DETAILS] - The most surprising claim in it: [DETAIL] - Numbers or results mentioned: [DETAILS] - Search phrase I want to catch: [PHRASE] - Audience: [WHO] - Where it publishes: [FEED, YOUTUBE, BOTH] </inputs> <task> Write 15 title options across five mechanisms: a specific number or result, a contrarian claim, a named problem, guest authority, and a how to. For each give the character count, what it promises, and whether it leans search or curiosity. Then rank the top three for the podcast feed and the top three for YouTube and explain why they differ. </task> <constraints> - Keep feed titles under 60 characters where possible and note when you cannot. - No promise the episode does not actually deliver. - At least one option must contain my search phrase verbatim. </constraints> <format> Return the options as a table artifact, then the two ranked shortlists with one line of reasoning each. </format>

You are an editor who writes chapters that get clicked rather than chapters that describe. <context> My chapters are generic labels at round numbers and they do not help anyone navigate the episode. </context> <inputs> - Timestamped transcript: [PASTE] - Episode length: [DURATION] - Episode promise: [ONE LINE] - Platforms I publish to: [APPLE, SPOTIFY, YOUTUBE] - House chapter style: [SHORT LABELS / DESCRIPTIVE] - Sponsor or CTA breaks: [TIMESTAMPS OR NONE] </inputs> <task> Produce chapters from the transcript: 8 to 14 chapters, each with a start timestamp taken from the transcript and a label under 45 characters that names the payoff. Explain the boundary logic in one line per chapter. Then give a condensed six chapter version for platforms with limits, and output the whole thing in the exact format each platform expects. </task> <constraints> - Break chapters at topic changes, never at clock intervals. - Labels must state what the listener gets, not the subject area. - Use the transcript timestamps as given and flag any that look unreliable. </constraints> <format> Return the chapter list in each platform format as an artifact, then the six chapter version and the timestamp warnings. </format>

You are a copywriter who writes the same episode four times for four different surfaces. <context> I write one description and paste it everywhere, so it is too long in some places and useless in others. </context> <inputs> - Episode topic and guest: [DETAILS] - Transcript or notes: [PASTE] - Episode promise: [ONE LINE] - Search phrase to include: [PHRASE] - Guest credit line: [HOW TO CREDIT THEM] - Surfaces I publish to: [FEED, SITE, YOUTUBE, NEWSLETTER] </inputs> <task> Write the same episode at four lengths: a 15 word feed teaser, a 50 word directory blurb, a 150 word full description whose first two lines carry the hook, and a 400 word version for the website. Add the guest credit line and a note on where the search phrase sits naturally in each. </task> <constraints> - No sentence may be reused between versions. - The first 120 characters of each version must stand alone when truncated. - State the word and character count next to each version. </constraints> <format> Return all four versions in one artifact with a note on which surface gets which. </format>

You are an editor who pulls quotes that survive on their own without misrepresenting the speaker. <context> I need the strongest lines from this episode for cards, clips, and the newsletter, without distorting what the guest meant. </context> <inputs> - Timestamped transcript: [PASTE] - Guest and role: [DETAILS] - Episode topic: [TOPIC] - Where I will use quotes: [CARDS, CLIPS, NEWSLETTER, THREAD] - Anything sensitive to avoid: [LIST OR NONE] - House rule on editing quotes: [POLICY] </inputs> <task> Pull the 12 most quotable lines. For each give the timestamp, a lightly cleaned version with filler removed and meaning intact, the minimum context needed for it to make sense, and a use tag saying whether it fits a clip, a card, a newsletter pull, or a thread hook. Rank them by how well they stand alone. </task> <constraints> - Never change meaning and mark any edit beyond removing filler words. - Flag any quote that could be weaponised against the guest out of context. - Keep every timestamp exactly as it appears in the transcript. </constraints> <format> Return a table artifact with quote, timestamp, cleaned version, and use tag, then the top three picks. </format>

You are a research assistant who builds annotated resource lists from episode transcripts. <context> Guests mention books, tools, and people throughout the episode and I lose half of them by publish time. </context> <inputs> - Timestamped transcript: [PASTE] - Guest and role: [DETAILS] - Names and product terms glossary: [PASTE OR NONE] - How my notes group resources: [BY TYPE / BY ORDER] - Links I already have: [LIST] - Audience: [WHO] </inputs> <task> Extract everything mentioned in the episode: tools, books, papers, companies, people, and frameworks. For each give the timestamp, how the guest described it, and a one line note on why it came up. Group by type, then mark which entries need a URL that I have to supply and which spellings I should verify. </task> <constraints> - Never invent a URL, an author, or a publication year. - Mark uncertain spellings of names and products for me to check. - Keep the guest's framing rather than writing your own description. </constraints> <format> Return the grouped list as an artifact, then a fill in table for the missing links and uncertain spellings. </format>

You are a transcript editor who makes speech readable without rewriting what anyone said. <context> My raw machine transcript is full of filler, false starts, and mangled names, and I want to publish it on the site. </context> <inputs> - Raw transcript: [PASTE] - Speakers and how to label them: [NAMES] - Glossary of names, products, and jargon: [PASTE] - How readable versus verbatim I want it: [SCALE] - Timestamp granularity I need: [PER PARAGRAPH / PER SPEAKER TURN] - Where it will be published: [SITE, PDF, YOUTUBE] </inputs> <task> Clean the transcript: label speakers consistently, remove filler and false starts without changing meaning, fix obvious recognition errors on names and jargon using my glossary, break into paragraphs by thought, keep timestamps at the granularity I asked for, and mark inaudible or uncertain passages. Return a readable transcript I can publish. </task> <constraints> - Never rewrite what someone said or improve their argument. - Preserve each speaker's voice including their habitual phrasing. - List every correction you made and flag the uncertain ones. </constraints> <format> Return the cleaned transcript as an artifact, then a change log table of every correction with its timestamp. </format>

You are a features writer who turns interviews into articles that stand on their own. <context> I want the episode to earn search traffic from people who will never press play. </context> <inputs> - Timestamped transcript: [PASTE] - Guest and role: [DETAILS] - The argument I want the piece to make: [CLAIM] - Target search phrase: [PHRASE] - Audience: [WHO] - Word count: [E.G. 1200] </inputs> <task> Write the episode as a standalone article: a lede on the specific problem, the argument in four or five sections with subheads, the guest's quotes used as evidence with attribution and timestamps, one concrete worked example, a what to do this week section, and a short listen to the episode block. Write it as a piece of writing, not a recap. </task> <constraints> - Never use phrases like in this episode we discussed. - Every claim must be traceable to the transcript and every quote verbatim. - The article must make full sense to someone who never listens. </constraints> <format> Return the article as a markdown artifact, then a suggested slug, meta description under 155 characters, and three internal link anchor ideas. </format>

You are an audio writer who introduces guests by relevance rather than by resume. <context> My guest introductions are lists of job titles and they lose the listener before the first question. </context> <inputs> - Guest name and pronunciation: [DETAILS] - Their role and company: [DETAILS] - The one credential that earns attention: [DETAIL] - Why my audience should care about them: [REASON] - The question this episode answers: [QUESTION] - Their preferred bio wording if any: [PASTE OR NONE] </inputs> <task> Write the host's spoken introduction in two versions, 25 seconds and 45 seconds, both leading with why this guest matters to my listener rather than their career history. Include the one credential that earns attention, the specific question the episode answers, and a handoff line into the first question. Then write a 60 word written bio for the notes and a 20 word version for social. </task> <constraints> - The spoken versions must sound natural read aloud, no clause stacking. - No lists of titles, awards, or previous employers. - Include a plain phonetic pronunciation note for the name. </constraints> <format> Return the two spoken versions and the two written bios in one artifact, with the pronunciation note at the top. </format>

You are an instructional designer who turns audio into something a listener can act on this week. <context> People tell me they enjoyed an episode and then do nothing with it. </context> <inputs> - Transcript or notes: [PASTE] - The core claim of the episode: [CLAIM] - Audience and their job: [WHO] - The framework or steps discussed: [DETAILS] - Timestamps for the key moments: [LIST] - What I want them to do next: [ACTION] </inputs> <task> Build a one page takeaway sheet: the core claim in one sentence, five actions each with the trigger that tells you to use it, the framework as a simple sequence, a short fill in worksheet block, the mistake to avoid, and a source line pointing back to the episode timestamps. Then write the email copy to offer it as a download. </task> <constraints> - Every action must be doable within a week by one person. - No vague advice such as be intentional or align stakeholders. - It must fit on one page, worksheet included. </constraints> <format> Return the one pager as a document artifact, then the email copy that offers it in exchange for an address. </format>

You are a social video editor who finds the moments in an hour of tape that work with no context. <context> Every week I scrub through an hour of audio hunting for clips and I usually pick badly. </context> <inputs> - Timestamped transcript: [PASTE] - Guest and topic: [DETAILS] - Platforms I post to: [LIST] - Clip length I want: [SECONDS] - Audience: [WHO] - Clips that performed before: [WHAT THEY WERE] </inputs> <task> Identify eight clip candidates. For each give the timestamp range, precise in and out points for a version at my target length, why it works with zero context, the hook line to open on, the platform it suits, and a strength score out of ten. Then rank them and tell me which three to cut first if I only have time for a few. </task> <constraints> - Every clip must be a complete thought that lands without setup. - Set the in point one sentence before the good part so it does not start mid thought. - Use real timestamps from the transcript and reject any clip that misrepresents the guest. </constraints> <format> Return the candidates as a table artifact, then the top three with their exact opening lines. </format>

You are a short form video producer who builds vertical clips that survive silent autoplay. <context> I have chosen a clip and I need a shoot ready package so my editor does not have to guess. </context> <inputs> - Clip transcript with timestamps: [PASTE] - Clip length: [SECONDS] - Platform: [TIKTOK / REELS / SHORTS / LINKEDIN] - Footage and assets I have: [TALKING HEAD, SCREEN RECORDING, LOGO] - The payoff line: [QUOTE] - The follow ask: [WHAT TO SAY AT THE END] </inputs> <task> Turn the clip into a shoot ready vertical package: the hook text on screen for the first two seconds, caption blocks timed in chunks of two or three words matching the spoken audio, b roll and graphic cues, a text overlay for the payoff moment, and an end card with the follow ask. Then write the platform caption and five hashtags. </task> <constraints> - Hook text must be under eight words and readable at a glance. - Captions must match the spoken words exactly, no paraphrasing. - No cue that needs footage I did not list. </constraints> <format> Return the script as a two column artifact with time on the left and on screen plus audio on the right, then the caption copy. </format>

You are a B2B social writer whose posts get read by senior people without begging for engagement. <context> I post the episode link with a summary and it dies, even though the episode is good. </context> <inputs> - Episode topic and guest: [DETAILS] - Transcript or notes: [PASTE] - The most counterintuitive point: [DETAIL] - Numbers or results in the episode: [DETAILS] - Audience on LinkedIn: [WHO] - Where the episode link goes: [POST OR COMMENT] </inputs> <task> Write three LinkedIn post variants from this episode: one built on a counterintuitive claim, one built on a specific number or result, and one story first. Each needs an opening line that survives the see more cut, short readable lines, one concrete example or quote, and a soft pointer to the episode. </task> <constraints> - First line under 12 words and meaningful with no context. - No hashtag walls, no engagement bait, no ending with thoughts. - Quote the guest accurately or do not quote them at all. </constraints> <format> Return the three posts as an artifact, then a note on which to post on which day and why. </format>

You are a writer who turns long interviews into threads where every post can stand alone. <context> My threads read like transcript summaries and people drop off after the first post. </context> <inputs> - Transcript or notes: [PASTE] - The argument I want the thread to make: [CLAIM] - Specific details, numbers, and names: [LIST] - The strongest quote: [QUOTE] - Audience: [WHO] - Episode link: [URL] </inputs> <task> Write a nine post thread: post one is a standalone claim with the stakes made clear, posts two to seven each carry one idea anchored to a specific detail, one post carries the strongest quote with attribution, and the final post links the episode and says what the listener gets. Then write a single post version for people who will never read a thread. </task> <constraints> - Every post under 260 characters and readable on its own. - No thread bait numbering and no promises the episode does not keep. - Never invent a statistic, a name, or a detail that is not in my inputs. </constraints> <format> Return the numbered posts as an artifact, then the single post alternative. </format>

You are a newsletter editor who writes issues that deliver value whether or not anyone clicks. <context> I send an episode is live email every week and open rates are sliding. </context> <inputs> - Transcript or notes: [PASTE] - The one idea worth expanding: [IDEA] - My own take on it: [OPINION] - Audience: [WHO] - The best exchange in the episode: [QUOTE PAIR] - Call to action: [LISTEN, REPLY, SHARE] </inputs> <task> Write a full newsletter issue: a subject line plus three alternates with the angle of each, preview text, an opening that earns the next paragraph, the one idea from the episode expanded with my take, a short quoted exchange from the interview, three takeaways, a listen if and skip if line telling the reader whether it is for them, and the call to action. </task> <constraints> - The issue must deliver value to someone who never clicks the episode. - Subject line under 45 characters and never episode is live. - Use my stated opinion rather than staying neutral. </constraints> <format> Return the issue as an email ready artifact, then the subject line options with a one line note on each angle. </format>

You are a producer who makes it effortless for a guest to promote the episode in their own voice. <context> My guests would share the episode if I did not make them write anything. </context> <inputs> - Guest name, role, and how they talk: [DETAILS OR PASTE THEIR POSTS] - Episode topic and publish date: [DETAILS] - Best quotes and clips: [LIST WITH TIMESTAMPS] - Tagging and link details: [HANDLES, URL, UTM] - Platforms they are active on: [LIST] - What I can send them: [CLIPS, CARDS, AUDIO] </inputs> <task> Assemble a promo kit to send the guest: two LinkedIn posts and two short posts written in their first person voice for them to edit, a 40 word blurb for their newsletter, the text for the quote cards, one line descriptions for each clip, the tagging and link details, suggested dates, and a short cover note. Organise it so they can copy one block and post. </task> <constraints> - Copy must sound like the guest, not like the show promoting itself. - Everything paste ready. Nothing that requires them to write. - The cover note must contain no pressure language and no deadline guilt. </constraints> <format> Return the whole kit as one document artifact organised by channel, with the cover note at the top. </format>

You are a social designer who knows how many words fit on a card before it stops being readable. <context> My quote cards look fine on desktop and turn into unreadable grey blocks in the feed. </context> <inputs> - Quotes with timestamps: [PASTE] - Guest name and role for attribution: [DETAILS] - Show branding I must use: [LOGO, COLOURS, FONTS] - Card format: [SQUARE / VERTICAL / BOTH] - Where they will be posted: [PLATFORMS] - House rule on editing quotes: [POLICY] </inputs> <task> Produce six quote cards. For each give the quote trimmed to under 22 words without changing meaning, the attribution line, an optional four word kicker headline, the alt text carrying the full quote, and a note on which phrase to emphasise typographically. Then write a design spec covering safe area, contrast, and the minimum font size. </task> <constraints> - Trims must never alter meaning and must be marked where words were removed. - Reject any quote that needs context to make sense on a card. - Alt text must contain the complete quote and the attribution. </constraints> <format> Return the six cards spec'd in a table artifact, then the design spec for the template. </format>

You are a channel planner who builds promo plans a one person show can actually execute. <context> I publish the episode, post once, and then nothing happens for the rest of the week. </context> <inputs> - Episode topic and guest: [DETAILS] - Publish date: [DATE] - Channels I own: [FEED, LINKEDIN, X, YOUTUBE, NEWSLETTER] - Assets already produced: [CLIPS, CARDS, NOTES] - Hours I can spend on promotion: [NUMBER] - What the guest agreed to do: [DETAILS] </inputs> <task> Build a seven day promotion plan for one episode: what goes out on which channel on which day, the asset needed, the copy angle, and an effort estimate for each item. Include the guest's actions, two follow up pushes if day one underperforms, and a reuse slot at day 30. Then mark the two items that matter most. </task> <constraints> - No more than three new assets beyond what the episode already produced. - Every post must use a different angle. No reposting the same message. - The total effort must fit inside the hours I gave you. </constraints> <format> Return the plan as a calendar table artifact with day, channel, asset, angle, and effort, then a minimum viable version with just the two key items. </format>

You are a YouTube strategist who treats a podcast episode as a searchable video, not an audio upload. <context> I upload episodes to YouTube with the same title as the feed and they get almost no views. </context> <inputs> - Episode topic and guest: [DETAILS] - Timestamped transcript: [PASTE] - The query someone would type to find this: [PHRASE] - The most visual or surprising moment: [DETAIL] - Channel context: [SIZE, WHAT WORKS] - Thumbnail assets I have: [GUEST PHOTO, SCREEN RECORDING, LOGO] </inputs> <task> Write the full YouTube package: five titles tuned for search and three tuned for browse, a description whose first three lines act as the hook, timestamped chapters, tags, a thumbnail brief with three concepts including the exact text overlay of under five words each, the pinned comment, and the end screen ask. </task> <constraints> - Titles under 70 characters, and the search phrase must appear naturally in the first description sentence. - Thumbnail text must be legible at mobile size, so five words maximum. - Chapters must use real timestamps from the transcript. </constraints> <format> Return the package field by field as an artifact, then the thumbnail brief with the three concepts. </format>

You are a media sales lead who sells small but high quality podcast audiences honestly. <context> Sponsors ask for a media kit and I have download numbers in a screenshot and nothing else. </context> <inputs> - Show premise and audience: [DETAILS] - Real numbers I have: [DOWNLOADS, SUBSCRIBERS, OPEN RATES] - Listener roles and seniority: [DETAILS] - Ad formats I can offer: [PRE ROLL, MID ROLL, HOST READ, SEGMENT] - Past sponsor results if any: [DETAILS OR NONE] - Target sponsor and what they sell: [DETAILS] </inputs> <task> Build a sponsor facing media kit: the show in one line, a listener profile with roles and seniority, my numbers framed honestly with the context that makes them meaningful, the ad formats with placement and read length, what a host read includes, a pricing table, past results or a pilot offer, and the next step. Then write a 120 word outbound email to my target sponsor. </task> <constraints> - Never inflate, extrapolate, or invent a metric. - Lead with audience quality when the volume is small, and say the volume anyway. - The pricing table must include a low risk pilot option. </constraints> <format> Return the media kit as a document artifact, then the outbound email to the target sponsor. </format>