Fewer subscriptions.
More vibe coding.

Get started

1. Get FFmpeg

   Use the official download page to choose a package or linked build for your operating system. Make sure ffmpeg is available in your terminal.

   ffmpeg -version

   Copy

2. Convert one file

   Place an input.avi file in your working folder and convert it to an MP4. Use a fresh output filename.

   ffmpeg -i input.avi output.mp4

   Copy

3. Save the recipe

   Ask your coding agent to wrap a working command in a small script with an input picker, progress output, and a separate export folder.

About FFmpeg

FFmpeg edits video and audio using commands. You can cut a recording into clips, resize it for different platforms, compress large files, convert formats, add subtitles, and create zoom or pan effects. Save those commands in a script and apply the same edits to a whole folder of videos.

• You supply the editing rules and zoom timing; automatic cursor tracking and choosing highlights require extra code.
• Long videos and complex filters can require substantial CPU or GPU time.
• License obligations depend on the build and codecs. Check the official licensing page before distributing FFmpeg inside a product.

Get started

1. Install the skills

   Run inside your project and choose your coding agent.

   npx skills add remotion-dev/skills

   Copy

2. Create a video project

   Use the official starter, then ask the agent to use remotion-best-practices.

   bun create video

   Copy

3. Preview and export

   Review the Studio preview. Ask the agent to render an MP4 when ready.

   npx remotion studio

   Copy

About Remotion

Create product videos, animated charts, captions, and social clips from code. Remotion's official skills help your AI agent build the video, preview it, and export the result. For example, give the agent a product name, screenshots, and a short script, then ask for a 15-second promotional video. Once the template works, change the inputs to make another version. This is especially useful for videos you produce repeatedly: weekly reports, product launches, or clips in several languages. You keep the project and can edit it with your agent. Remotion's license determines whether the runtime is free for your organization; agent usage and rendering can also cost money.

• The standalone skills repository does not publish a separate license grant; the linked license belongs to the upstream Remotion project.
• Remotion is source-available, with free-use eligibility limits. Internal use does not automatically exempt larger companies.
• You supply media, an AI agent, and rendering resources. This does not capture your screen or replace every video-editor feature.

Get started

1. Get the self-hosted stack

   Install Docker and follow the official Docker guide's manual installation steps to copy a tagged self-hosted release into a new project folder.

2. Configure the instance

   Follow the guide to generate secrets and signing keys, set your app and API URLs, and configure email. Replace all example passwords before startup.

3. Start and connect

   From the configured project folder, start the stack. Open Studio, create a table, enable row-level security, and connect your app with the appropriate client key.

   sh run.sh start

   Copy

About Supabase

Supabase gives your app a Postgres database, user accounts, file uploads, and APIs for reading and changing data. The self-hosted version runs on your own server. You can use it to build a customer portal, inventory tracker, or booking app without writing the login and database services yourself.

• You are responsible for security patches, backups, monitoring, and recovery.
• Self-hosted services do not provide the full managed Cloud platform or its support guarantees.
• Email delivery and SMS authentication need separately configured providers and may add costs.

Get started

1. Install the browser CLI

   Install the CLI and its browser runtime.

   npm install -g agent-browser && agent-browser install

   Copy

2. Install the skill

   Choose your coding agent in the installer.

   npx skills add vercel-labs/agent-browser

   Copy

3. Load current instructions

   The skill asks the agent to read this before using the browser.

   agent-browser skills get core --full

   Copy

About Agent Browser

Give your AI agent a browser so it can click buttons, fill forms, collect information, and take screenshots. The Agent Browser skill shows it how to use Vercel's browser automation tool on your computer. You could ask it to collect product names and prices from a list of pages, download reports from your own dashboard, or check that a signup form works. The same instructions can become a small tool you run again next week. You provide the agent and the computer running the browser. For larger jobs, you also need to handle scheduling, login sessions, failures, and any websites that restrict automated access.

• Local execution needs a running computer; scheduling, retries, and monitoring are your responsibility.
• Logins, changing page layouts, and site restrictions can interrupt automation.
• It provides browser control, not a hosted crawler network or guaranteed extraction accuracy.

Get started

1. Get the local extension

   Download Build.zip from the repository's releases page and unzip it.

2. Load it in Chrome

   Open chrome://extensions, enable Developer mode, and choose Load unpacked. Select the extracted build folder.

3. Record and export

   Record a short walkthrough, review it in the editor, and export a video. Start your custom workflow from that file.

About Screenity

Screenity records your screen, camera, and microphone from Chrome. You can draw arrows or notes while explaining something, trim the recording, and export a video file. Use it for product demos, tutorials, and bug reports. The open-source extension can run locally without a recording subscription.

• The extension depends on Chrome's capture permissions and browser capabilities.
• Self-hosted builds are local-only. Hosted link sharing and Screenity Pro features are separate.
• It is not a complete replacement for Screen Studio's automatic presentation effects or a full video editor.

Get started

1. Install the skill

   Choose the agent you use for the project.

   npx skills add anthropics/skills --skill frontend-design

   Copy

2. Provide a real brief

   Share the audience, page goal, copy, assets, and framework. Ask for frontend-design explicitly.

3. Build and review

   Have the agent implement the page, then inspect mobile layout, focus states, and the main interaction.

About Frontend Design

Build a landing page, dashboard, or small website by describing what you need to your coding agent. Anthropic's Frontend Design skill guides the agent through choosing a layout, fonts, colors, and useful interactions, then writing the page in your project. You can supply your own copy, screenshots, and brand assets so the result fits the product. For example, ask for a simple site to launch an internal tool, or a clearer screen for managing invoices. You keep the code and can host it yourself. The skill helps with the design and implementation work; forms, payments, content management, and hosting still need to be set up separately.

• The skill is guidance, not a drag-and-drop editor, CMS, or hosting platform.
• Generated designs and code still need review; accessibility and responsive behavior are not guaranteed.
• Stock assets, fonts, third-party packages, and hosting have their own terms.

Get started

1. Get the source

   Install Git, CMake, and a C/C++ compiler, then clone the project and enter its directory.

   git clone https://github.com/ggml-org/whisper.cpp.git
   cd whisper.cpp

   Copy

2. Download a model and build

   This small English model is a useful first test. Use a multilingual model for other languages.

   sh ./models/download-ggml-model.sh base.en
   cmake -B build
   cmake --build build -j --config Release

   Copy

3. Transcribe the sample

   Run the included sample, then check the CLI help before processing your own audio. Start with a short WAV file and check which input formats your build supports.

   ./build/bin/whisper-cli -m models/ggml-base.en.bin -f samples/jfk.wav

   Copy

About whisper.cpp

whisper.cpp turns recorded speech into text using a Whisper model running on your computer. Give it an interview, voice note, or video soundtrack and it produces a transcript. It can also export timed subtitle files. You download the model once and provide the processing power instead of buying transcription minutes.

• Transcripts need review, especially for names, accents, noisy audio, and silence.
• Larger models need more memory and can run slowly without suitable acceleration.
• Meeting bots, reliable speaker identification, and summaries are separate features to build or add.

Get started

1. Install the skill

   Add the upstream testing instructions to your coding agent.

   npx skills add anthropics/skills --skill webapp-testing

   Copy

2. Prepare Playwright

   Use a Python environment, install Playwright, then download Chromium.

   python -m pip install playwright && python -m playwright install chromium

   Copy

3. Test one flow

   Give the agent your dev-server command and a specific journey; ask it to use the bundled server helper and keep the test.

About Webapp Testing

Ask your agent to open your web app and check that important actions work. Anthropic's Webapp Testing skill uses Playwright to click through pages, fill forms, take screenshots, and capture browser errors. You could test a signup form with missing fields, confirm that search returns the right results, or check a checkout using test payment details. The agent can save a script so you can repeat the checks after changing the app. This gives a small project a practical way to catch broken user flows before release. You run the tests on your own computer or CI service and maintain them as the interface changes.

• It runs the browsers and environment you provide, without a hosted device farm.
• Tests still need stable fixtures, sensible assertions, and maintenance when the UI changes.
• Agent-generated coverage is limited to the flows you request and verify.

Get started

1. Install the local app

   Download Ollama for your operating system from the official site and start it. Choose a local model rather than a cloud model.

2. Try a suitable model

   Choose a model from the official library that fits your available memory. Read its model card and license, then run a short task in the app.

3. Connect your tool

   Use the documented local API at http://localhost:11434. Keep the chosen model name in configuration and check the server is available before submitting jobs.

About Ollama

Ollama downloads and runs language models on your computer. Your own app can send text to a local model and receive a summary, draft reply, or suggested category. Use it to add basic AI features to internal tools without paying a hosted provider for every request.

• The MIT license covers Ollama, not every model you download. Check the selected model's terms.
• Model quality and speed depend on hardware and model size; local models may struggle with complex tasks.
• Local processing applies only when you select local models. Cloud features send requests off the machine.

Get started

1. Install the skill

   Select the coding agent that can create and inspect local files.

   npx skills add anthropics/skills --skill canvas-design

   Copy

2. Define the output

   Supply dimensions, exact text, brand colors, and images you can use.

3. Export and inspect

   Ask for a PNG or PDF plus reusable source, then check the final render for clipped text and margins.

About Canvas Design

Create a poster, announcement graphic, or report cover from a written brief. Anthropic's Canvas Design skill helps your agent arrange text, shapes, colors, and your own images into a finished PNG or PDF. Give it the dimensions, exact wording, and brand assets, then ask for a design you can export. It works well for a focused job such as an event poster or a cover you update every month. Ask the agent to keep the source code so later changes are easy to repeat. You still need to review the final image for readable text, correct details, and spacing before publishing or sending it to print.

• This produces files through an agent, not an interactive collaborative design editor.
• The skill does not include a stock-image library; supplied assets and fonts retain their own licenses.
• Check print dimensions, color requirements, and text legibility yourself before production.

Get started

1. Install Community Edition

   Use the official self-hosting guide to get the Community Edition binary for your operating system. Pin a stable release when deploying.

2. Configure and start

   Set a strong master key and persistent data location, then start the engine using the self-hosting guide. Keep administrative credentials out of browser code.

3. Index a small dataset

   Import a JSON document list through the API. Wait for indexing to finish, test a few searches, and connect a search-only key to your interface.

About Meilisearch

Meilisearch adds fast search to a website or app. Give it product records, article text, or document descriptions, and it returns matching results as someone types. It handles misspellings and lets you add filters such as category, price, or date. The open-source Community Edition can run on your own server.

• You must build and maintain synchronization from your source database to the search index.
• Enterprise capabilities such as sharding are outside the MIT Community Edition scope.
• Self-hosting requires capacity planning, backups, access controls, and upgrades.

Get started

1. Install the skill

   Install the complete folder so the Python utilities are included.

   npx skills add anthropics/skills --skill slack-gif-creator

   Copy

2. Install its dependencies

   In a Python environment, install the packages listed by upstream.

   python -m pip install pillow imageio imageio-ffmpeg numpy

   Copy

3. Generate and validate

   Give the agent an asset and motion brief; ask it to use GIFBuilder and the bundled validator.

About Slack GIF Creator

Make a small animated GIF from an icon, logo, or original drawing. Anthropic's Slack GIF Creator skill gives your agent Python tools for building animation frames, controlling movement, and keeping the file small. You could turn a team mascot into a bouncing reaction emoji, animate a product icon, or make a short loop for a message. Give the agent your image and describe the motion, then ask it to export and check the GIF. Keep the generating script to create more versions with different colors or images. This handles simple, short animations; detailed character animation and long videos need a more capable editing workflow.

• Best suited to small, short animations; GIF has limited color and compression quality.
• Slack upload limits can change, so verify the actual upload as well as the bundled checks.
• Python dependencies and image assets have their own licenses; the skill does not provide an asset library.

Get started

1. Start a local instance

   With Docker installed, run the official all-in-one image with persistent storage. This example binds the web interface to this machine only.

   docker run -d --name baserow -e BASEROW_PUBLIC_URL=http://localhost -v baserow_data:/baserow/data -p 127.0.0.1:80:80 baserow/baserow:2.4.0

   Copy

2. Create a working table

   Open http://localhost, create your account and workspace, and import a small CSV or define the fields for one workflow.

3. Connect an automation

   Create a database token with only the required table permissions. Use the generated API documentation to read a row before adding write operations.

About Baserow

Baserow gives your team shared tables for organizing records such as leads, equipment, and content ideas. People can edit the rows in a browser, while your scripts read or update the same records through an API. The open-source edition can run on your own server.

• Premium and Enterprise features are separately licensed; self-hosting does not unlock them.
• Moving from another table tool may require rebuilding formulas, views, and integrations.
• You manage the database, file storage, upgrades, and backups.

Get started

1. Install the skill

   Add the server-building workflow to your coding agent.

   npx skills add anthropics/skills --skill mcp-builder

   Copy

2. Scope the connector

   Provide the API documentation and start with a few read-only operations. Have the agent follow the TypeScript or Python reference.

3. Verify the server

   Use MCP Inspector to exercise the generated tools before connecting your agent.

   npx @modelcontextprotocol/inspector

   Copy

About MCP Builder

Let your AI agent look up information or perform tasks in systems you already use. Anthropic's MCP Builder skill helps it write a connector between the agent and an API, such as your inventory service or internal customer database. The connector exposes clear actions like finding a product or checking stock, so you can request them in a conversation. Start with a few read-only actions and test them before adding changes to data. This can replace a specific connector or a small integration you would otherwise build in an automation service. You still need API access, somewhere to run the connector, and an agent that supports MCP.

• MCP supplies agent-facing tools; it does not provide scheduling or a no-code workflow dashboard.
• External APIs still require permission, credentials, and any applicable paid plan.
• You maintain authentication, deployment, monitoring, and safeguards for any write operations.

Get started

1. Choose Community Edition

   Follow the official Docker guide. Use the appsmith-ce image in its Compose example and pin a release tag; the default example uses the separate commercial edition.

2. Run and connect

   Start the configured Docker Compose stack, create an administrator account, and add a read-only database or API connection first.

   docker compose up -d

   Copy

3. Build one screen

   Add a table widget and connect it to a query. Introduce a form and a confirmed update action only after the read path works.

About Appsmith

Appsmith helps you build internal admin pages using ready-made tables, forms, buttons, and charts. Connect those components to your database or an API, then add queries and JavaScript for the actions you need. Its open-source Community Edition runs on your own server.

• Commercial features such as advanced governance are outside Community Edition; check your requirements first.
• It is optimized for internal tools, so a highly custom consumer interface may need a separate frontend.
• The all-in-one deployment includes supporting services and needs more server resources than a static site.

Get started

1. Install the official skill

   Select your coding agent in the installer.

   npx skills add slidevjs/slidev

   Copy

2. Create a slide project

   Use a current Node.js environment and follow the starter prompts.

   npm init slidev@latest

   Copy

3. Build and preview

   Ask your agent to edit slides.md using the skill, then review the deck in the browser.

   npm run dev

   Copy

About Slidev

Turn a brief, document, or dataset into a presentation your agent can edit as code. Slidev's official skill covers slide layouts, diagrams, animations, code examples, and speaker notes. You can ask for a product demo, a workshop deck, or a monthly update using your existing screenshots and numbers. Preview the slides in a browser, then export a PDF or build a website to share the interactive version. The content lives in Markdown files, which makes it easy to update the same deck next month or reuse the design for another topic. You provide the AI agent and review the slides before presenting; exports have some format limitations.

• Requires a JavaScript project and agent; it does not include a managed collaborative deck workspace.
• Standard PPTX export uses slide images. Editable PPTX export has layout limits and can fall back to images for unsupported content.
• Interactive browser features do not fully survive file export; CLI exports also require playwright-chromium.

Get started

1. Get the Docker configuration

   Install Docker and Git, then clone the official repository. Review the supplied Compose configuration and use a stable release for deployment.

   git clone https://github.com/umami-software/umami.git
   cd umami

   Copy

2. Start your local instance

   The supplied Compose file runs Umami with PostgreSQL. Open http://localhost:3000 and change the default administrator password immediately.

   docker compose up -d

   Copy

3. Track one site

   Add your website in the dashboard and install its generated tracking script. Trigger one test visit and one named event, then verify both appear.

About Umami

Umami shows how people use your website. Add its tracking script to collect visits and traffic sources, then record events for actions such as clicking Download or completing signup. You get a dashboard for the results and can run the application with your own database.

• You maintain the server and PostgreSQL database, including backups and retention.
• Browser blocking and failed network requests can make analytics counts incomplete.
• Complex product analytics or attribution workflows need their own evaluation; basic event tracking is the comparison here.

Get started

1. Install the skill

   Choose your coding agent and add the JSON Canvas instructions.

   npx skills add kepano/obsidian-skills --skill json-canvas

   Copy

2. Generate a canvas

   Supply your notes or process. Ask for a .canvas file with groups, labeled connections, and validated node IDs.

3. Open and review

   Place the file in an Obsidian vault, or open it in another JSON Canvas-compatible app. Check spacing, references, and connections.

About JSON Canvas

Turn notes or a written process into a visual board with cards and connecting lines. Kepano's JSON Canvas skill helps your agent create mind maps, flowcharts, and project overviews as .canvas files. You could map a customer onboarding process, group research links by topic, or show how project tasks depend on each other. Open the result in Obsidian or another compatible app to move cards around and make changes yourself. The file format is open, so your board remains readable data that you can keep and edit. The skill generates the local board; shared editing, access permissions, and online collaboration depend on the app you choose.

• The skill and JSON Canvas format are open; Obsidian is a separate application with its own license.
• This does not include real-time collaboration, team permissions, or a hosted board-sharing service.
• A compatible viewer is required. Linked files and embedded content must remain accessible when you move or share the board.

Get started

1. Install Node-RED

   Install a supported Node.js version and use the official local installation guide. With a user-owned npm global directory, install the package below.

   npm install -g node-red

   Copy

2. Start the editor

   Start the runtime and open http://localhost:1880. Keep it on a trusted local network until you configure the documented access controls.

   node-red

   Copy

3. Build a first flow

   Connect an Inject node to a Function node and then a Debug node. Deploy, inspect the output, and replace the sample logic with one real API call.

About Node-RED

Node-RED lets you connect tasks in a visual editor: receive data, change its format, and send it to another service. A task can start on a schedule, from a webhook, or when a device sends a message. Add small JavaScript functions when you need custom rules. It runs on your computer or server.

• The editor is not secured by default; configure authentication and HTTPS before exposing it remotely.
• Third-party nodes vary in maintenance, and API connectors may need custom configuration or code.
• Retries, duplicate prevention, monitoring, and external API limits remain your responsibility.

Get started

1. Install the skill

   Select SEO Audit and your coding agent.

   npx skills add coreyhaines31/marketingskills --skill seo-audit

   Copy

2. Supply the evidence

   Provide the site URL, business goal, target pages, and any Search Console or crawler exports.

3. Audit and verify

   Ask for evidence-linked findings; verify each issue before changing the site and recheck affected pages afterward.

About SEO Audit

Check your website for problems that can make its pages harder to find in search. Corey Haines's SEO Audit skill gives your agent a checklist for reviewing page titles, links, indexing settings, content, and other technical details. Share your site, its source code, and any Search Console reports, then ask for the most useful fixes with evidence for each one. Your agent can also make the changes in the project and check the affected pages again. This is useful before a launch or after moving a website. It covers the review and repair work; keyword volumes, backlink databases, and ongoing rank tracking still require separate data or tools.

• It does not provide proprietary search-volume, backlink, or ranking datasets.
• Rendered pages and Search Console evidence may be needed; static HTML alone can miss client-side behavior.
• Findings depend on the evidence supplied, and recommendations do not guarantee rankings.