Video skills for your AI agents
Introducing video skills for product teams, customer education, and L&D teams who need to create and edit videos for internal training, customer education, and product launches.
Think of the last time you used AI for work. Now think about the last time you used it to create or edit a video. How long did that take? Were you happy with the final output?
It probably took several rounds of back and forth, 32 cups of coffee and maybe some braincells to get something passable. Right?
However, on October 16, 2025, Anthropic announced the release of Agent Skills, a new way to help Claude tackle complex tasks by providing more context and clearer instructions.
Context setting is a painful and demanding job.
Before agents, most of your output was limited to how you prompted a model. The idea was to front-load as much context as possible and hope the LLM incorporated every specification in your prompt without missing a beat.
There were two problems with this approach:
You had to load context every time you wanted an LLM to work how you wanted.
Bloated prompts with lots of context-setting would cause performance issues or simply confuse the LLM, ultimately affecting the final output.
What a skill actually is
A skill is a written set of instructions an AI agent reads before it starts working. It carries the decisions an experienced person would make.
For instance, what a compliance training video needs to contain, how long an opening hook should run, where captions belong are all part of the skill so the agent using it produces something usable on the first pass instead of something generic.
The useful consequence is that the knowledge moves. A skill called crm-training-video holds the production judgment, which means the person who knows your CRM doesn't also have to know what a good training video looks like. They describe the workflow. The structure comes from the skill.
Introducing video skills
Video creation—whether it’s internal trainings, product launch videos, or customer education material—has notoriously been rigorous and demanding. You create one set of videos only to watch them go stale as soon as your product goes through a UI/UX overhaul.
We created 90 skills spanning customer education, L&D, and product marketing use cases. You can start by exploring what these skills can do here.
Skills for L&D teams
L&D teams rarely need one video. They usually need 100, in six languages, all built the same way, produced by regional trainers who each have their own idea of what a training video should be.
The usual answer to maintain consistency here would be a style guide. But at such scale it hardly works because not everyone would read your style guide before recording their screen. A skill enforces rules that the agents have to follow. Every trainer who uses course-module-video for instance, produces the same chaptering, the same intro treatment, and the same caption behavior — without knowing that's what they're doing.
The multilingual piece is where the time actually goes. batch-multi-language-export and compliance-training-localization handle the version sprawl that makes a 40-video library into a 240-video library the moment you add five languages.
Skills for customer education
Customer education has a different shape of problem. The library isn't too small, it's too old. The product ships weekly, and the content team is just two people.
The skill that matters most here is article-to-video. Most customer education teams have already written the help center article — that's the artifact that gets updated when the product changes, because it's cheap to update. Making the video derive from the article inverts the usual order, where the video is a separate project that quietly falls out of sync with the docs.
annotate-article-screenshots and chapterize-video cover the rest of the loop, and add-ai-captions handles the accessibility pass that otherwise gets skipped under deadline.
We also created a skill for anyone documenting a product with real customer data on screen: blur-sensitive-info. Teams burn hours redacting account names and PII from screenshots by hand, then redo it on every product update and again for every language. Hopefully that’s not your reality anymore after using this skill.
Skills for product marketing
Product marketing's constraint is the queue. A launch video is a project that needs the creative team, and the creative team has capacity for the tier-1 releases. Everything below that ships with a changelog entry and no video.
brief-to-launch-video is built for exactly that gap — the release that deserves a video but will never get scheduled for one. Pair it with branded-intro-outro, apply-brand-colors, and add-cta-end-card and the brand consistency that normally requires a designer per asset comes from the skill instead.
You can now keep making awesome product launch videos and eleven other releases that will help you carry the quarter.
How to setup video skills
We know that the people who will end up using these skills aren’t developers so we spent time making sure setting these skills up is easy.
There's no terminal, no API keys, and nothing to install locally. You open your LLM’s settings, add a custom connector, and paste one URL: https://connect.clueso.io/mcp. Then you copy a skill from the site into your chat and ask for a video. A paid Claude plan is required for custom connectors.
Agent tooling is mostly distributed through channels built for engineers, which quietly excludes the L&D manager and the customer education lead who have the most repetitive video work in the company.
What does this mean for the future of video production?
These skills are built to replace the tedious work that’s part of video production. They don't remove your editorial judgement and taste from the equation.
Start by solving issues like adding voiceovers in different languages, generating and syncing captions to the narration in a video, and fixing pronunciation issues in existing videos. The first draft would still need a review pass, and rough source recordings would still produce rough results. But you’ll get a first draft with the structure and technical details that look good, which removes a review cycle rather than removing the reviewer.
Anyone promising zero-effort video is selling something. We don’t do that. We want to minimize your efforts when it comes to tedious work.
Start with a video you've been putting off
Every team has a video that's been on the list for two quarters. Maybe it's the onboarding module that's three UI versions out of date, or the release that shipped with a changelog and no walkthrough. Pick that one.
Add https://connect.clueso.io/mcp as a custom connector, copy article-to-video into your chat, and point it at a help center article you've already written. The first draft follows the structure you decided on when you wrote the article. What's left is deciding whether it says the right thing.
All 90 skills are on videoskills.dev, and they're open source under Apache-2.0.
You can fork them and rewrite the parts that don't match how your team works. If you build a better version of one, we'd like to see it.
Get Started with Clueso
Create stunning videos and documentation with AI today.
Senior Content Marketing Manager
Ashish is a Senior Content Marketing Manager at Clueso with 7+ years of experience across content, product, brand marketing. Now his mission is to help product and customer education teams realize the value of video-based learning. Outside of work, Ashish sketches, sings, plays the guitar, cooks, and does all things LLMs can't yet.
