Skip to content

gooseworks-ai/goose-skills / skills/video/write-video-ad-script

write-video-ad-script

Verified

Write or revise a short-form ad from a sourced angle and reference. Plan connected words, product proof and sound; expose production gaps before script review.

references/human-behavior.md

View on GitHub
# Human speech and behavior

**Summary.** Write the people, then the lines. Each speaker needs a reason to speak, a
limited view of the situation and a response to the person in front of them. An ad can
sell openly and still sound human. This method is a hypothesis to test on complete
scripts; correct grammar, a naturalness score and factual claims do not establish success.

## Give the writer usable evidence

Use the brand message, product facts and selected commercial interaction from
the `ad-angle-miner` skill and [dialogue-writing](dialogue-writing.md). Add a small
matched set of language evidence when the project lacks it:

- Public social discussions: retain the question or parent, reply, source URL and medium.
  A lone sentence loses the reason it was said. Reddit, X and LinkedIn writing can supply
  concrete concerns and response behavior; it is not automatically spoken language.
- Recorded conversation or captions: inspect an exchange, not an isolated hook. Captions
  establish reported words, not delivery you have listened to. Editorial speech may teach
  a language mechanic but cannot replace the required complete commercial street reference.
- Human-approved and rejected script examples: keep the exact words and the human's
  specific reason. Label newly written illustrations as unapproved. Do not label public
  posts human-authored or successful ads without evidence.

Reuse a relevant source library. Fetch a gap in role, concern or response behavior;
do not fetch unrelated trending content or copy a source's biography into the participant.
Language evidence never establishes product efficacy, customer identity or endorsement.

## Draft from the encounter

Use this instruction with the brief and references:

> Establish where the people are, what they were doing, why contact begins, what each
> wants immediately and what each knows. Separate the advertiser's message from the
> participant's purpose. A participant can agree to a sponsored interview without
> becoming the sponsor's copywriter. Use ordinary stakes appropriate to their role.
>
> Write each turn from what that person just heard or saw. Ask for something the asker
> actually wants to know. Let the reply change the next turn: an answer, clarification,
> correction, recognition or joke is enough. Do not add conflict or quirks by quota.
>
> Write the words someone would use to accomplish that immediate purpose. Use complete,
> understandable questions rather than headings, briefing notes or a menu of concepts.
> Short replies are allowed when they answer the actual question. Naturalness is not
> a requirement to omit grammar, add filler or avoid promotion.
>
> Establish why a recommendation happens here. An advertiser can identify their role
> and introduce the offering directly. A participant can ask about it after noticing
> something relevant. A complaint alone does not invite a prescription. Do not invent
> a convenient question solely to unlock a sales paragraph.
>
> Explain one relevant product detail in ordinary words. Preserve the supported meaning
> without reading the product description aloud. If a specification has no useful place
> in the exchange, move it to a permitted card, demonstrative action or explicit address
> to viewers. The participant cannot know unintroduced brand facts or claim a result
> they have not experienced. Do not invent product use, professional status or endorsement.
>
> Read the exchange aloud in the situation. Fix the cause of a false-feeling line before
> changing its wording. Then deliver the requested complete scripts: opening action,
> exact dialogue and ending. A critique or research report does not replace the scripts.

## Contrast the actual failure with the intended behavior

These are **manager-rejected lines**, not approved examples. The repairs below describe
what to change; they are not a bank of replacement lines to reuse across brands.

| Rejected words | Failure to repair |
| --- | --- |
| “I keep replaying the shot I missed.” | The writer assigned an athlete's consuming concern to an ordinary recreational player without earning it. Reconsider the ordinary person's situation before polishing the sentence. |
| “What makes you put one back?” | The object and decision are unclear. Ask the actual question this encounter raises using words this person would say. |
| “Before bed: another episode, or twenty minutes to wind down?” | A prompt menu was put in the interviewer's mouth. Choose an ordinary question with a clear subject and purpose. |
| “GooseWorks has AI coworkers that can make those photos using your brand's notes.” | A feature paragraph was inserted where a person should respond. Establish the speaker's role, reason to introduce GooseWorks and what the listener needs explained. |

A better-looking sentence can still fail. Do not defend a rejected line by inventing
context, professional identity or unusual behavior outside the actual scene.

## Review the complete script

For each line identify its immediate cause and what the speaker is trying to do.
If the line could follow any reply, reconsider the exchange. If a recommendation depends
on private intentions, put its social context into the actual action or words. Review
the literal viewer opening as well as the whole scene; private notes cannot fix either.

Preserve human rejection when a model disagrees. Keep timing and claims as separate checks.
Model scores screen the frozen words; the human reviews those same words. Openly selling
is allowed. The target is believable people, not an ad disguised as a conversation.

## Research basis and limits

[John August's dialogue article](https://johnaugust.com/2007/how-to-write-dialogue) models
listening before replying and revises an information dump into a locally triggered reply.
[Writing Excuses](https://writingexcuses.com/17-31-everyone-has-an-agenda/) distinguishes
the author's purpose from each character's purpose. These are craft guidance, not
experiments proving that a prompt produces good ads.

The October 2026 project research collected public Reddit and LinkedIn parent/reply
exchanges, a LinkedIn video transcript and X posts through ScrapeCreators. It found
concrete concerns and replies reacting to details, but did not verify human authorship,
ad performance or a prompt that reliably produces human-sounding street ads. Keep source
records and limits in the project; test this instruction on complete new scripts.
receiving-code-reviewobra/superpowersUse when receiving code review feedback, before implementing suggestions, especially if feedback seems unclear or technically questionable - requires technical rigor and verification, not performative agreement or blind implementationStars: 322kUpdated Claude CodeCodexVerifiedInstall
requesting-code-reviewobra/superpowersUse when completing tasks, implementing major features, or before merging to verify work meets requirementsStars: 322kUpdated Claude CodeCodexVerifiedInstall
using-git-worktreesobra/superpowersUse when starting feature work that needs isolation from current workspace or before executing implementation plans - ensures an isolated workspace exists via native tools or git worktree fallbackStars: 322kUpdated Claude CodeCodexVerifiedInstall
verification-before-completionobra/superpowersUse when about to claim work is complete, fixed, or passing, before committing or creating PRs - requires running verification commands and confirming output before making any success claims; evidence before assertions alwaysStars: 322kUpdated Claude CodeCodexVerifiedInstall