Claude Code Skills: Five Public Packs, What They Actually Do, and What to Check First

On this page · 9 sections
A skill is a saved pack of instructions a coding helper can load when a job matches. In most cases it's a folder with a SKILL.md file, and sometimes scripts. The helper loads the relevant skill when you ask it to do that kind of work, or when you call the skill by name.
Claude Code is Anthropic's coding app in the terminal. Several of these packs also work with other helpers, including Codex and Cursor. "Free" in this post means the public files. The helper itself is often a paid product, and a long job still burns tokens (the chunks of text the model reads and writes).
We're AI Lab Australia, a Sydney AI agency. We set up coding agents for teams. We have not installed these five packs on a client's repository. The facts below come from the public repos, official pages and skill files, which we re-read on 12 November 2026. Treat the numbers as a snapshot.
Matt Pocock's engineering skills
Matt Pocock's pack is a set of composable engineering skills: specs, tickets, test-driven development (a failing test before the code that makes it pass), code review, triage and architecture. The repo is mattpocock/skills. The homepage is aihero.dev/skills. The repo was created on 3 February 2026. The licence is MIT, a common free software licence that allows commercial use.
At our 12 November 2026 check, there were 20 skills under skills/engineering/, and 39 SKILL.md files in the repo as a whole, including productivity skills and in-progress work. The pack works with more than one coding helper, including Codex and Cursor. It is listed in Anthropic's Claude plugin marketplace, so you can install it from Anthropic's official plugin list. The files are still Pocock's, and the other four packs in this post are not on that list.
The official Claude Code install command, from the project's README, is:
claude plugin install mattpocock-skills@claude-plugins-official
A skill pack like this can nudge an agent toward better habits. You still have to read the work.
Emil Kowalski's apple-design
apple-design lives in Emil Kowalski's wider skills pack, alongside other interface skills such as mobile-native. The repo is emilkowalski/skills, created on 17 March 2026, Sydney time. The skill file is skills/apple-design/SKILL.md. The homepage is aiforui.dev/skills. The licence is MIT.
The skill translates Apple WWDC design talks for the web. WWDC is Apple's developer conference. The skill file says it draws chiefly on Designing Fluid Interfaces (WWDC 2018), typography from WWDC 2020, and eight "Principles of Great Design" from WWDC 2026. The topics include response, springs, momentum, interruptible motion, depth, typography and restraint. One of the 2026 principles is written as "Simplicity, not minimalism."
This is not a dump of "over 50" official Apple principles. Apple's 2026 list, as the skill itself records it, is eight principles. The skill cannot ensure a clean, minimal look. It is a different project from dickwu/apple-design-skill, which reviews work against Human Interface Guidelines pages. None of these is an official Apple product.
The install command in the repo README is:
npx skills@latest add emilkowalski/skills
Humanizer
Humanizer is a skill that marks and rewrites common signs of AI writing. Siqi Chen published it under the GitHub account blader/humanizer. The repo was created on 18 January 2026. The licence is MIT.
The pattern list comes from Wikipedia's Signs of AI writing guide, which WikiProject AI Cleanup maintains. The README currently lists 26 patterns. It is meant to keep the meaning of the original text and not invent facts. Ars Technica covered an earlier version in January 2026, when the list had 24 patterns.
Humanizer cannot catch every sign of machine writing. A later model will invent new habits. The current README is also blunt about detectors: Humanizer "does not try to get past AI detectors, and detectors still flag most of its output." It will not make a false claim true. If the draft is wrong, a tidier sentence is still wrong.
Lemo-Opuscar
Lemo-Opuscar, by Lemomo / LemoLab, is a library of film styles plus a skill that directs a short film in code. The repo is lemomo-ai/lemo-opuscar. The gallery is at lemomo-ai.github.io/lemo-opuscar. The repo was created on 26 September 2026.
The agent draws frames with canvas and WebGL (web graphics libraries), adds text-to-speech narration, and writes music from free sample libraries. The README says there is no video-generation model. Official wording on 12 November 2026 is 99 ready-to-use styles, after 33 new styles were added that day. A count of STYLE.md files in the repo on the same day was 100. An earlier public figure was 66, and both the English and Chinese READMEs now list 99.
The styles were tuned with Claude Opus 5.5. The current README says Claude Fable, Opus and Sonnet all work, and to set effort to high (in Claude Code, type /effort high). Other models may not reproduce the styles well. Some styles need an image generator you provide. First-run tools include Node 20 or newer, ffmpeg 7 or newer, and Python 3.11 or newer. The README says a film takes an agent about 30 to 60 minutes and a fair amount of tokens.
The licence is the part people miss. It is not a single MIT grant. GitHub's licence field on the repo is "Other". The LICENSE file splits it:
- Code is MIT.
- Docs and style guides, including each style's STYLE.md, are Creative Commons Attribution-NonCommercial 4.0 (CC BY-NC 4.0). That licence blocks most commercial reuse of the docs themselves. On top of that, the author gives a limited extra permission: you and the people working with you on a film may read, follow, copy and adapt the docs to make your own films, including commercial and client films, and may include those docs in the project files you deliver for that film. Selling the docs, or putting them into a commercial product or service, needs a commercial licence from the author.
- Demo films and stills are mostly Creative Commons Attribution 4.0 (CC BY 4.0), with named exceptions, including some narration tracks, Anthropic's Clawd character, and third-party marks.
- Films you make with the library are yours, commercial films included.
This is general information, not legal advice. If you plan to wrap the style guides into a paid product, read the LICENSE and ask the author. If you want a short film for a client, the extra permission is written for that job, with the limits above.
The official install commands in the README are:
claude plugin marketplace add lemomo-ai/lemo-opuscar
claude plugin install lemo-opuscar@lemolab
Corey Haines' marketing skills
Corey Haines' pack is a collection of marketing skills for coding helpers. The repo is coreyhaines31/marketingskills. The homepage is marketing-skills.com. The repo was created on 16 January 2026. The licence is MIT.
At our 12 November 2026 check, the live skills/ directory and the homepage both listed 53 skills, including search engine optimisation (seo-audit, ai-seo), copywriting, ads (ads, ad-creative) and email (emails, cold-email). Installing all 53 at once loads a large instruction set. Pick the skill that matches the task.
The README's recommended install command is:
npx skills add coreyhaines31/marketingskills
What "free" does not cover, and what to check first
Four of these packs are MIT-licensed. Lemo-Opuscar is the exception, and it is a real one for any Australian business that wants to put the style guides into a commercial product. "Free" still never covers the model bill or the computer tools. Lemo-Opuscar also needs Node, ffmpeg and Python on the machine that runs the job.
For a team in Australia, the boring checks come before the install command.
Read the skill file before you load it. Look for instructions that fetch remote content, send data away, or change files outside the project. Read any hooks or install scripts, because those run code on your computer. Get the address from the official GitHub repo, and check the owner and repo name character by character. We have not independently audited these five packs for malware.
If the helper will see client files or personal information, the Privacy Act 1988 can apply. The Office of the Australian Information Commissioner (OAIC) says the Act covers AI tools that touch personal information. Our Privacy Act guide covers this. Keep client data out of a trial until you know where the model runs and what the vendor's terms do with prompts. If the work cannot leave Australia, the agent setup is as important as the skill; see private AI in Australia.
We wrote a longer install-hygiene checklist in our post on Ponytail and how to install agent skills safely. For tools that scan skills before you enable them, see our roundup of open source AI security tools for AI agents. A skill is a set of instructions. It is not a review process, a test suite or a lawyer.
FAQ
What is a Claude Code skill?
A skill is a saved pack of instructions, usually a SKILL.md file and sometimes scripts, that a coding helper such as Claude Code can load for a particular kind of job.
Are these five packs official Anthropic or Apple products?
No. Matt Pocock's pack is listed in Anthropic's official plugin marketplace, so you can install it from that list. The files are still his. apple-design is a third-party translation of public WWDC talks, not an Apple product.
Are the skill packs free to use?
The public files for four of the five packs are under the MIT licence. "Free" means those files, not the coding helper or the model usage. Lemo-Opuscar's code is MIT, but its docs and style guides are CC BY-NC 4.0, with a limited extra permission to use the docs to make your own films, including commercial films. Selling the docs, or putting them into a commercial product, needs a commercial licence from the author.
Can I use Lemo-Opuscar on a client film?
The LICENSE gives you and the people working with you on a film permission to use the docs to make your own films, commercial and client films included, and to put those docs in the project files you deliver for that film. It does not cover selling the docs or putting them into a commercial product or service. Read the LICENSE, and get legal advice if the use is unusual. This is general information, not legal advice.
Can Humanizer make AI-written text undetectable?
No. The current README lists 26 patterns, says Humanizer does not try to get past AI detectors, and says detectors still flag most of its output. It also will not make a false claim true.
How should I install a skill pack safely?
Use the official GitHub repo or the official install command in that repo's README. Read the skill file and any scripts before you enable it. Don't install a pack from a forwarded link. Run the helper away from production credentials, and review the output.
Will these counts stay the same?
Probably not. On 12 November 2026 we counted 20 engineering skills and 39 SKILL.md files in Pocock's repo, 26 patterns in Humanizer, 99 Lemo-Opuscar styles (100 STYLE.md files), and 53 marketing skills. Lemo's public style count was 66 earlier.
Want a second look at the skills your team already uses?
If you want a straight answer on which skill packs to allow, email info@ailabaustralia.com with the helpers and skills your team uses. We'll tell you what we'd keep, what we'd read first, and what we'd leave off a client machine.
Email info@ailabaustralia.com →
Sources
- Matt Pocock, skills and aihero.dev/skills, checked 12 November 2026.
- Anthropic, Matt Pocock's Skills marketplace listing, checked 12 November 2026.
- Emil Kowalski, skills, apple-design SKILL.md and aiforui.dev/skills, checked 12 November 2026.
- Siqi Chen, Humanizer, checked 12 November 2026.
- Wikipedia, Signs of AI writing, and WikiProject AI Cleanup, checked 12 November 2026.
- Benj Edwards, Ars Technica, 21 January 2026.
- Lemomo / LemoLab, Lemo-Opuscar, LICENSE and gallery, checked 12 November 2026.
- Creative Commons, Attribution-NonCommercial 4.0.
- Corey Haines, marketingskills and marketing-skills.com, checked 12 November 2026.

WRITTEN BY
Roman Silantev
CEO and Lead AI Engineer. Scopes and builds AI systems for Australian businesses.
← PREVIOUS9 Open Source RAG Tools for Australian Businesses (2026)Keep reading in Tools & models
All articles →
Ponytail for Claude Code: What 'Half the Tokens' Really Measures, and How to Install Agent Skills Safely10 MIN READ
Subframe Motion for Australian React Teams: 310 Free Animations Your Coding Agent Can Search9 MIN READ
Removing Smart TV Ads with Claude Code: CleanYourTV and Free My TV, Checked for Australian Homes and Offices8 MIN READ
