CommunityWriting & Editinggithub.com

builtnotboosted/video-scripting

Writes video scripts that sound like the person who will say them, for shorts (Reels, TikTok, YouTube Shorts) and long-form YouTube. Warm hook rather than a shout, who-am-I and follow early, a retention shape that delays the answer, a soft earned close, length that follows the story, real-problem substance, reactive scripts for news and data releases. Output is teleprompter-ready (think-breaks, shot reminders) with short edit notes for the video-editing skill. Use when someone says "write me a script", "script a reel about", "I need a short on", "plan a YouTube video", "turn this post into a video", "react to today's news on video", "what should I say on camera", or before a filming session.

What is video-scripting?

video-scripting is a Claude Code agent skill that writes video scripts that sound like the person who will say them, for shorts (Reels, TikTok, YouTube Shorts) and long-form YouTube. Warm hook rather than a shout, who-am-I and follow early, a retention shape that delays the answer, a soft earned close, length that follows the story, real-problem substance, reactive scripts for news and data releases. Output is teleprompter-ready (think-breaks, shot reminders) with short edit notes for the video-editing skill. Use when someone says "write me a script", "script a reel about", "I need a short on", "plan a YouTube video", "turn this post into a video", "react to today's news on video", "what should I say on camera", or before a filming session.

Works with~Claude Code~Codex CLI~Cursor
npx skills add https://github.com/builtnotboosted/built-not-boosted-skills/tree/HEAD/plugins/built-not-boosted/skills/video-scripting

Installed? Explore more Writing & Editing skills: steipete/notion, langchain-ai/langchain, bytedance/podcast-generation · View all 6 →

Ask in your favorite AI

Open a new chat with this agent skill pre-loaded.

Documentation

Video scripting

This skill owns the spoken script and its edit notes. It writes what the person says on camera, in their voice, shaped to hold attention, and hands three short lines to the video-editing skill so the edit knows the on-screen hook, the cover words and the promised outcome. Captions and post copy belong to the writing skill, format choice and topic research to research, the edit itself to video-editing, and QA to review-loop.

The core belief: a script is a skeleton for a real person, not a read-aloud. People who sound good on camera explain around the page. Write for that, and the video sounds like them instead of like an advert.

Before you start

  1. Find the marketing workspace (default marketing/). If none, say so in one line and offer brand-setup, or ask the essentials: who is speaking, what they do, who for, and where.
  2. Read brand/brand-profile.md: ## Identity, ## Audience, ## Voice, ## Boundaries, ## Compliance, ## CTA & contact. Brand-profile wins on brand constants and compliance wording; this skill owns the method.
  3. Read brand/voice-learnings.md first, every script. It is the record of how this person actually talks on camera, learned by the voice-loop skill from their filmed videos and hand-edits. Apply its current working rules; they beat the general guidance here wherever the two differ.
  4. Read brand/performance-learnings.md for what has worked (formats, hooks, lengths) if it exists.
  5. Get the facts. Any claim, figure or quote comes from the person, the brief or research with a source. Never invent a customer story, a figure, a quote or a review. If there is no real story, make it honest education instead.

The standard

  • Warm and relaxed, not frantic. The opposite of hard-cut, high-energy shouting. Inviting, unhurried, human. Light humour where it lands naturally, never at a customer's expense.
  • Sounds like them. If a line would not come out of their mouth, rewrite it. Plain words, contractions, varied sentence length.
  • Stand beside the viewer, not in front of them. Avoid "most people get this wrong" and "people often think", which tell the viewer they are probably wrong. Say "it's easy to miss", "a lot of us were told". Avoid hollow hype signposts ("here's the bit nobody tells you") that make the viewer wait for a payoff; introduce the real point plainly.
  • Useful, never a sell. Build trust and the following; let enquiries come from that.
  • One idea per short. Real problems your audience actually has, not fluff.
  • No em dashes in script text. Commas, brackets, full stops.

Structure: shorts (30 seconds to about 2 minutes, always under 3)

A shape, not a formula. Many creators will recognise it as Hook, Intrigue, Deliver, Exit; we do it the warm way.

  1. Hook. One full spoken sentence, roughly 10 to 14 words, that names the topic plainly and gives the human stake or the honest twist. Not a staged scene, not a riddle, not "hi guys", not a logo. The tension lives in the fact, not in a hyped delivery. Write it as the whole sentence they would say (many people naturally open with a run-in like "So,"); clipped fragment questions tend to get rewritten on camera.
  2. Who-am-I and follow, straight after. One short beat in the person's own words: name, what they do, for whom, and where if local, then a light follow ask. Example shape: "I'm Sam, I fix bikes at the workshop on the high street. Give me a follow and let's get into it." Keep the identity line consistent across videos (that is the point of it); vary only the tail so a batch filmed in one afternoon does not open identically. Put the location here, not in the hook. Never an outcome promise, never begging ("smash that follow").
  3. Intrigue. One or two sentences, often from their real week ("had someone in this week who..."), that make the viewer feel the problem. Its job is to delay the answer, never to give it.
  4. Deliver. The one clear idea, unhurried, in plain words. Pay off exactly what the hook promised; if the hook numbers or withholds something, write the payoff first and make sure the body cashes it in.
  5. Exit. One small thing to try today, with a question about it: "Grab a ruler and check your chain tonight. How did yours measure? Let me know in the comments." One thread, one ask, phrased as "this helped me", never "you must". Then a quick, human goodbye on its own line, in their words (keep it the same every time once you know what they say).

No-spoiler rule: neither the hook nor the first line after the who-am-I may give the answer away. If it does, the viewer has what they came for and scrolls on.

Structure: long-form (about 8 to 15 minutes, 16:9)

  1. Warm open: the hook (the promise of the video), a light who-am-I for new viewers, and if you like a soft subscribe reason ("if this is useful, stick around, there's more like it").
  2. Setup: frame the problem and say what the viewer will get.
  3. Chapters: each a mini setup, teach, payoff. Make every transition a small hook forward rather than a stopping point, and plan two deliberate re-engagement beats around the one-third and two-thirds marks. Chapter markers in the script are silent markers for the edit; viewer-facing chapters come from the description timestamps.
  4. Payoff and ending: resolve the promise, then exactly three asks: one thing to try, a pointer to a related video of theirs, and one comment question. Then a proper close.
  5. Writing order: payoffs first, then setups, then tension, hook last, so the hook honestly matches what the video delivers.
  6. Search: title and description lead with the viewer's question in their words ("why did my boiler pressure drop?") rather than the mechanism.

For reactive scripts (a news story, a price change, a data release), and more hook patterns and retention detail, open references/hooks-and-retention.md.

Length and pace

  • Length follows the story. Do not pad to a runtime and do not cut a real explanation to hit 40 seconds. Short shorts (30 to 60 seconds) and fuller 1 to 2 minute pieces are both fine when the story is told.
  • Size by pace. Until voice-learnings holds the person's measured pace, assume about 190 words a minute: about 95 words for 30 seconds, 190 for a minute, 380 for two. Deliberate one-point pieces run slower (around 150), brisk round-ups faster (around 220). About 500 delivered words is a practical ceiling for a short.
  • Write close to the spoken length. Early on, some people expand a script to twice its length on camera; once the delivered-to-written ratio in voice-learnings settles near 1.0, write at full length. Let the voice-loop numbers decide.
  • Facts in full, opinions as bullets. People read facts and quotes almost word for word but riff on opinion beats, so write the checkable facts and any quote in full and leave opinion beats as short prompts.
  • Write plainer than feels satisfying on the page. Ornamental phrases get replaced on camera with the plain version. Save the craft for one homely metaphor and the payoff line; concrete, everyday images survive filming best.

Compliance

Brand-profile ## Compliance decides. If the business is regulated or the video promotes something with required wording, the default is that the required warning rides in the caption (written by writing), not spoken, so the script stays human and the caption carries the exact words. Check your own rules: some regimes or ad policies require the warning on screen or spoken, and then brand-profile says so and you follow it. Either way, keep tension in real facts, never promise an outcome ("you'll definitely save"), and do not hard-quote a price or figure on camera that may change before it airs. A labelled example for UK financial services is in references/regulated-industries-example.md.

Output (one folder per piece)

In content/<yyyy-mm-dd>-<slug>/:

  1. script.txt, teleprompter-ready: spoken text only, clean line breaks, no headings, no em dashes. Think-breaks (a blank line every two or three sentences and before the payoff) so the speaker can glance away and come back, which also gives the edit clean cut points. Three to six shot reminders on their own lines in angle brackets and capitals, two or three words each, just before the line they apply to: <WALK IN>, <STEP CLOSER>, <HOLD COVER LOOK>, <B-ROLL: KEYS, DOOR>. Details and the PromptSmart example in references/prompter-and-edit-notes.md.
  2. edit-notes.txt, three lines, never on the prompter:
    HOOK ON SCREEN: Gears slipping?
    COVER: CHECK THIS FIRST
    OUTCOME: you'll know the two usual causes of skipping gears and how to check one tonight
    
    HOOK ON SCREEN is 3 to 5 words that promise without giving the answer. COVER is 2 to 4 true, plain words for the thumbnail. OUTCOME is what the viewer gets; if you would not want it as a viewer, do not make the video.
  3. sources.txt: where every fact came from.
  4. Baseline copy: save an untouched copy of script.txt to baselines/<yyyy-mm-dd>/<slug>-script.txt. Never edit it. The voice-loop skill diffs the person's edits and their filmed delivery against it, which is how the next script gets closer to them.

Captions and titles for the posts are then written by writing from the script and edit notes.

Checks before it ships

Pass the script and edit notes to review-loop, or check: hook lands in the first seconds and is not a spoiler; who-am-I and follow early, no outcome promise; intrigue delays the answer; every promise in the hook is paid off; one try-this step with a question about it; quick goodbye; no "most people", no hype signposts; on-screen hook is 3 to 5 words and not the answer; think-breaks and shot reminders present; length sized to the pace; no invented story or figure; compliance per brand-profile; voice-learnings rules applied; no em dashes.

Adapting it

  • Your words: put the person's who-am-I line and goodbye in brand-profile ## Voice or ## CTA & contact, and let voice-loop refine them from real footage.
  • More than one presenter: give each their own who-am-I and keep their voice-learnings separate if they talk differently.
  • Your prompter: any teleprompter app works; check once that its voice tracking scrolls past shot-reminder lines cleanly.
  • Your boundaries: anything you cannot advise on or claim goes in brand-profile ## Boundaries; scripts can educate about it and signpost, never position you as doing it.

References

  • references/hooks-and-retention.md: hook patterns, on-screen hooks, retention shape, reactive and data-release scripts, closes. Open when the hook is not landing or for a news reaction.
  • references/prompter-and-edit-notes.md: teleprompter formatting, shot reminders, edit notes and shoot habits. Open when preparing a filming session.
  • references/worked-example-short.md: a full short for a neutral small business, script plus edit notes.
  • references/regulated-industries-example.md: how this works for a regulated business (UK financial services as the example).

Learn more: https://builtnotboosted.com/skills/video-scripting

Individual skills in this repo

This repo contains 4 individual skills — each has its own dedicated page.

builtnotboosted/image-generation

Produces finished social images with any image model (Google Gemini image, "Nano Banana", as the worked example) - feed statics, carousel slides, short-form video covers, YouTube and article thumbnails, local listing cards and animated statics. Bakes the headline into the scene instead of pasting text on top, keeps text inside the profile-grid safe zone, locks the likeness of real people to their reference photos, hits exact platform sizes, writes alt text and re-judges every re-roll on the filed file. Use when someone says "make an image for this post", "create a thumbnail", "make a cover from my video", "design a carousel", "animate this static", "the image model won't do 4:5", "my headline is getting cut off in the grid", or when a content pack needs its visuals.

builtnotboosted/review-loop

The quality gate every piece of content passes before it is filed, scheduled or posted. Runs a self-critique, then an independent review by a fresh subagent that sees only the finished work and the criteria (never the builder's reasoning), then fix and re-judge until there is a clean pass. Covers what to do when a piece can't pass unattended (hold, flag, ship the rest). Use when someone says "review this", "check this before it goes out", "QA the pack", "is this ready to post", "get a second pair of eyes on this", "proofread the captions", or when any other Built Not Boosted skill hands work over for checks.

builtnotboosted/scheduling

Gets finished posts into a social media scheduler (or posted by hand) safely. Queue-only with explicit timezone offsets, live dedup against scheduled and already-published posts, media readiness checks, correct post types (shorts, reels, carousels, first comments, manual-posting reminders), alt text, a link gate that sends copy back for a rewrite rather than stripping it, and confirmation before anything goes live. Writes and keeps a scheduling manifest for each pack. Works with any scheduler; Publer is the worked example. Use when someone says "schedule this", "queue these posts", "put this in Publer", "book in this week's content", "post this on Tuesday at 7", "move that post", "what's already scheduled", or hands over a finished content pack to go out.

builtnotboosted/video-editing

Edits filmed talking-head footage into release-ready vertical shorts (1080x1920) and 16:9 YouTube long-forms with ffmpeg and Whisper. Keeps exactly one clean read of every script line (no stumbles, restarts or double reads), never leaves a flat single shot (punch-ins, reframes, B-roll, motion graphics), captions from the finished audio, normalises loudness to -14 LUFS, makes a real-frame cover and runs a QA gate that re-transcribes the render and refuses delivery until it passes. Syncs extra cameras by sound and writes YouTube chapters. Use when someone says "edit my video", "cut this clip", "remove my retakes", "edit out the fluffs", "add captions to my video", "make a short from this footage", "sync my two cameras", "make a thumbnail from my video", or hands over raw footage and a script.

Related Skills