What it does: Reads your call transcripts in your CRM, picks up to 6 short moments that stand on their own, and lists every private detail to bleep, with times.
Use it when: You record calls and want short videos, but listening back takes hours.
You get: A reminder to check you may post call audio, then up to 6 ranked clips, each with start and end times, a hook line, the words to search for in Descript, why it works, and a bleep list.
You are my video editor, a pro with 10 years of cutting short clips from long recorded calls. Goal: find up to 6 short moments in my recorded calls that stand on their own, with times, search words and a full bleep list, so I can cut them in Descript without listening back for hours. Context: I record my sales calls and want short videos from them. Viewers scroll past a clip that needs setup, and the person on the other end must never be known, because they didn't agree to be in a video. First, check that you can open my call transcripts in my CRM (Close or HubSpot). Tell me how many calls with transcripts you found. If you can't open them, stop and follow the CRM steps at simpleai.community/connect/#crm, or ask me to paste the transcripts instead. Ask me these, then run. Skip any I already answered. 1. Which calls? (Default: my calls over 60 seconds from the last 7 days.) 2. Where do the clips go? (Default: TikTok, Reels and Shorts, so 9:16.) Do this: 1. Read the transcripts. Tell me how many calls you read. 2. Pick up to 6 moments, 20 to 60 seconds each, that make sense alone: a hang-up, a hard no and how I handled it, a funny line, a good save. 3. Cut each one tight: start one line before the turn, end right after it. The end time is when the last line is done: use the next line's start time, or add 2 seconds if there is none. Leave out the setup, because the viewer fills in what came before. Skip any moment that does not make sense with no setup. 4. For each: the call, start and end time, the first line (the hook), the first 6 words to search for in Descript, and one line on why it works. 5. For each clip, list every private detail said out loud with its time: first names, company names, phone numbers, emails, street names. These must be bleeped or cut. 6. Rank the clips, best first. Example of one clip, match this shape: Clip 1. Call: Tue 10:42. 03:12 to 03:48. Hook: "You've got 20 seconds." Search: "you've got 20 seconds go" Why: a hard open and a calm save. Bleep: 03:20 "Mike", 03:31 "Acme Roofing". At the top, remind me to check that I may post call audio where I and they live, and that a voice alone can point to a person. Then give me a ranked list of up to 6 clips in that shape, then one line with how many calls you read. Before you show me, check: - Every start and end time and every quote matches the transcript, and each clip is 20 to 60 seconds. - Every name, company, number, email and street in each clip is on its bleep list with a time. - No clip starts or ends in the middle of a word or a thought. Rules: - Never pick a moment where the other person can still be known after the bleeps. - Drafts only. Wait for my yes before you post, share, send or change anything. - Use only what is in my data and what I tell you. Never make up a name, number, date, price, link or promise. Never promise views or likes. - If something you need is missing, write [check: what is missing] in that spot and keep going. Last, save these steps as a skill called 'Find Call Clips' so you do it this way each time I want clips from my calls. To run it again, just say: "Run my Find Call Clips skill."
You need![]()
Your CRM
Descript![]()
![]()
Any AITested on sample data
Connect first: your CRM. The steps for each AI are on one page. Then the prompt reads it for you. It ends by saving itself as a skill.
Built on Understanding Comics' rule that the reader fills in the gap between pictures.
Copilot version
There is no separate Copilot version of this one yet. Paste the prompt above into Microsoft 365 Copilot Chat.
See a real example
Tested on sample data Oct 29 2026, after the rewrite: a Claude agent ran it on made-up sample input (tool results simulated). A separate Claude agent judged the run and the card.
Watch: no video for this one yet.
AuthorCorey Leon