Ready To Record VSL Scripts . Agency OS

What it does

Produces a finished video sales letter script you can read into a camera or a teleprompter today. Every script carries timing markers every 30 to 60 seconds, labelled sections, slide notes or camera direction depending on format, psychological triggers annotated inline, and a word count against a 150 words per minute pace. It picks from 8 frameworks (Jon Benson, Russell Brunson, Alex Hormozi, Stefan Georgi, Jim Edwards, Cole Gordon, PAS Extended and Mini VSL) based on your price point and offer type. After you approve the script it delivers the production package: slide outline or teleprompter version, three short ad cuts, and five landing page headlines.

Say this to start

This skill has no button. You start it by saying what you want. Any of these will do it:

/write-vsl

> write a VSL script

> video sales letter

> sales video script

> talking head script

> video pitch

When to reach for it

Writing a VSL or sales video script
A talking head or text-on-screen sales video
An application page video to pre-qualify high-ticket prospects
A short lead magnet video that sells the opt-in
A condensed webinar pitch in video form
A video pitch for a client's product in their voice

When NOT to use it

If you actually want	Use this instead
The pages, opt-ins and email or SMS sequences that surround the VSL	`funnel-architect`
Non-video persuasion assets: emails, sales pages, long-form sales letters	`legendary-copywriters`
The paid ad copy driving traffic to the VSL page	`meta-ads-copywriter`
Captions and posts for a video you already recorded	`social-caption-writer`
The shot list, effects plan or B-roll behind the video	`video-prompt-builder`
Thumbnails and slide visuals	`creative-director`
Slicing an already-recorded video into clips and posts	`video-repurposer`

Before you start

What you need	Why
`epistemology.md` in the workspace	This is a hard blocker. The skill refuses to write without it and tells you to run profile deepening first, because the problem reframe, the old way versus new way structure and the diagnostic walk-through are all epistemology patterns. Without them you get a swap-the-noun script	Required
A defined target audience	Every line of the script targets one person's frustration. Build the avatar first if you do not have one	Required
The product or service and its price	The framework recommendation keys directly off the price band, from under 100 dollars up to 5,000 and above. No price, no framework choice	Required
The desired CTA: buy, book a call, apply, watch next	The whole close is built around one action, and the skill treats multiple CTAs as a conversion killer	Required
`brand/[workspace]/verified-claims.md` with entries marked `consent: ok-as-stated`	A VSL is proof-dense by design. Any result, testimonial, credential or dated story that is not in the registry gets genericised and flagged as a proof gap	Required
Proof points: testimonials, case studies, data, credentials	One of the six quality gates checks that proof appears in every major section, not just one block. Thin proof shows up as gaps throughout the script	Optional
A guarantee or risk reversal	The close leans on it. Without one the offer section carries more objection weight	Optional
`ica-output.md` and `offer-output.md`	They feed pain points, language patterns, objections, value stack and price anchoring straight in. Without them the skill asks you the same questions manually	Optional

How it runs

Choose format, then stopIt presents three formats: text-on-screen for fast low-production turnaround, talking head for personal brands and high-trust offers, hybrid for premium or complex products. It recommends two or three frameworks based on your price band, offer type and audience, explains the reasoning, and waits for you to confirm.
Gather script intelligenceIt auto-loads your ICA, offer and voice profile from the workspace. Anything missing it asks you directly: audience and their frustration, what you sell and what makes it different, price and payment plans, proof, guarantee, and CTA.
Generate the scriptIt loads the chosen framework reference and the cross-framework persuasion principles, then writes the full script with timing markers, labelled sections, psychological triggers in braces, estimated length and word count. Text-on-screen scripts add slide count and per-slide timing. Talking head scripts add energy markers and camera direction.
Review section by section, then stopIt presents the script in five blocks: hook, problem and story, solution and mechanism, offer and value stack, close and CTA. It confirms each meets the framework requirements and offers three alternative hooks to test. It waits for your feedback before finalising.
Run the quality gatesSix checks: does the hook stop someone in five seconds, does it match the audience's awareness level, is proof in every major section rather than one dump, are the top five objections handled, is there exactly one unmistakable CTA, and does the emotional arc move through fear, hope, excitement and urgency in sequence.
Run the gradersA VSL is heard verbatim, so the hook and every hook variation go through a hook grader, and the full script plus the ICP go through a copy grader. Weak hooks get rewritten and any axis scoring below 4 gets resolved. The claim grader governs proof separately.
Deliver the production packageOnce the script is approved you get the slide outline or the clean teleprompter version, three ad cuts at 15, 30 and 60 seconds pulled from the strongest sections, and five landing page headline options derived from the hook and core promise.
Log and learnIt writes a vsl-output schema artifact so downstream skills can consume the script, appends the VSL to assets.md, and asks whether you shipped it as-is, made minor edits or rewrote it. That answer is logged to learnings.md and shapes the next script.

What you get

A complete script with timing markers every 30 to 60 seconds and labelled sections
Slide notes and visual cues in parentheses, psychological triggers annotated inline in braces
Estimated total runtime, word count and words per minute target
A slide-by-slide outline with exact text and timing, for text-on-screen format
A clean teleprompter version with no annotations, plus energy and camera notes, for talking head format
Three video ad cuts at 15, 30 and 60 seconds pulled from the strongest sections
Five landing page headline options derived from the hook and core promise
Three alternative hooks to test
A vsl-output schema artifact, and a log entry in brand/[workspace]/assets.md
Deliverables as separate markdown files named in snake_case

Honest limits

Read this before you rely on it

It refuses to run without epistemology.md. This is the only skill of the four that hard-blocks rather than warning and continuing. If you have not deepened the profile, the skill will send you to do that first.
Specifics are capped by your registry. A VSL demands proof in every section, so a thin verified-claims.md shows up as repeated [PROOF GAP: ask the client for a real number/result here] flags through the script rather than as invented results. That is the gate working, and it is the clearest signal of which proof you need to go and collect.
The claim checks do not see chat. The linter fires on file writes to client-facing files, so a script you read straight out of the conversation has passed the model grader only. Read the proof gap flags before recording.
It is confirmation-gated, so it is not one-shot. There are two hard stops where it waits for you: after the framework recommendation and after the script review. Budget attention for both.
It writes the script, nothing else, and you supply the commercial facts. No filming, no editing, no voiceover, no slide design, no thumbnails, and no funnel pages or email sequences around the video. It will not invent your price, your proof or your guarantee, and the framework recommendation itself depends on you knowing your price point.

Where people go wrong

The mistake	Do this instead
Opening the script with credentials, or asking the skill to	Nobody cares who you are yet. Open with a result, a pattern interrupt or a bold claim about their situation. Credentials land later, after you have proved you understand the problem.
Revealing the price before the value stack is built	Build the full stack first and anchor against the cost of the problem and the price of alternatives. Then the price reads as a fraction rather than as the headline number.
Stacking all the proof into one testimonials section	Distribute it: one proof point after the hook, one in the story, one in the offer, one in the close. Viewers who drop before your proof block never see any of it.
Offering the viewer several next steps	One CTA, repeated two or three times in the close. Save any secondary action for the email follow-up.
Arriving without a price point	Decide the price first. The framework recommendation is keyed to price bands, so without one the skill cannot make its main decision.
Approving the script and skipping the production package	The package is where the reusable assets are: three ad cuts and five page headlines, both derived from the script you already approved.

Worth knowing

Bring the price and the real proof to Step 1, not Step 3. The framework choice is decided almost entirely by price band, and the six quality gates are decided by proof density, so those two inputs shape the script more than any brief you write. Arriving with a confirmed price and a filled registry turns the first confirmation stop into a decision instead of a discovery session.