agentleFS
inreels-aiinreels-ai/inreels-skills/skills/product-demo-ad/SKILL.md

product-demo-ad

Skill1 starsChanged 12 days agoMITView on GitHub

Make a product-led video ad with no presenter — the product doing its job, in motion, with native audio. Use when asked for a product video, demo ad, product-only ad, or when a product is visual enough that a talking creator gets in the way.

The file

We split this skill into the rules it gives an agent

3 rules. About 0 are common, 3 are distinctive, and 0 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈3

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 0

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    No face. The product does the work, the camera stays close, and the viewer sees the thing happen.

    line 8
  2. Distinctive:

    Only props the action needs.

    line 71
  3. Distinctive:

    A cluttered product image: Every frame inherits it. If the label is illegible at thumbnail size, the ad is illegible in the feed.

    line 111

≈ “Common” is an estimate: at least 1 in 5 of the 372,429 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Product demo ad

No face. The product does the work, the camera stays close, and the viewer sees the thing happen. For a product that demonstrates itself this beats a creator ad, and it costs the same.

When to take this over ugc-ad

product-demo-ad          ugc-ad
──────────────────────────────────────────────
it visibly DOES something it needs vouching for
the result is on screen   the result is internal
B2B / tools / hardware    beauty / supplements / apparel
no likeness questions     a face builds the trust

A blender, a cleaning spray, a label maker, a software UI: demo. A vitamin, a mattress, a course: creator.

Requirements

inReels MCP connected — get_account to verify. create_ad_video needs a paid plan.

A product image is required. Omit creator_actor_id and creator_image_url and you get the product-led ad.

Method

1. Find the one action worth watching

The ad is built around a single physical thing happening. Name it before you write a word of script:

  • cup onto the base → press → it blends
  • cap off → pump → rub in, and the texture changes
  • the strip peels and the surface is clean behind it

An appliance that stays off is a dead ad. We have rendered this and it is boring in exactly the way you expect. If the product has an on state, turn it on. If it has a before and after, show both in one continuous move.

2. Write for the action, not over it

Narration explains what is not visible. The screen handles what is.

LengthWords
5sup to 10
10sup to 20
15sup to 30

Over budget and the line gets clipped mid-sentence — the model does not speak faster to fit. Cut words rather than hoping.

Often the best product demo has no narration at all: the sound of the thing working, and one line of text at the end. If you want that, say so in the brief and keep the script short or absent.

3. One product, one surface

The model juggles objects badly. Hold the scene down:

  • Exactly one of the product in frame, never a duplicate.
  • Only props the action needs. Nothing arrives that was not there at the start — a surprise second cup or a straw appearing from nowhere is the giveaway that this was generated.
  • The product stays in one place or one hand through the move.

4. Quote, then render

quote_ad_video { "product_image_url": "…", "script": "…", "variations": 3 }
create_ad_video {
  "product_image_url": "…",
  "script": "<within budget, or omit for brief>",
  "brief": "<the action, if you are letting the model write the line>",
  "look": "natural",
  "aspect_ratio": "9:16",
  "variations": 3
}

look: commercial is the one case where it earns its place — a product-led ad is allowed to look produced. natural still reads as more trustworthy on Meta. Test both if the budget allows; that is what variations is for.

5. Poll and hand over

get_run_status { "run_id": "…", "wait": true }

Five to ten minutes. watch_url appears before the MP4 does — it is a preview page, not the deliverable. Hand over output_url, and only once the run reports completed.

What makes these fail

  • The product never does anything. The single biggest one. Static product on a turntable is a stock video, not an ad.
  • A cluttered product image. Every frame inherits it. If the label is illegible at thumbnail size, the ad is illegible in the feed.
  • Describing the product by brand name. On-screen brand text comes out misspelled. Describe shape, colour and parts instead.
  • Narration that reads the screen aloud. Say what the picture cannot.
  • Too much action in a short take. A 5s ad holds one move. Asking for three gets you cuts you did not want.

Spending rules

  • create_ad_video spends credits and needs a paid plan.
  • quote_ad_video first; the quote is exactly what the call charges.
  • Each variation is charged separately — quote the set and state the total.
  • Failed runs refund in full.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.