openai / plugins

accessibility-and-inclusive-visualization

A chart that still gets its point across to someone who cannot see it the usual way — read aloud by a screen reader, driven by keyboard, without colour, or with the animation turned off.

agent-browser

A browser you drive from the command line — open a page, click, type, screenshot — so an agent can work a website the way a person would.

agent-browser-verify

You just started a local web server; this opens the page in a real browser and checks that it actually renders, instead of trusting that the server said it started.

agents-sdk

Turns an idea into a small working app built on OpenAI's Agents SDK, runs it on your own machine to prove it works, and can put it online when you ask.

ai-elements

Ready-made pieces of a chat interface — message bubbles, streaming text, tool-call panels — so you assemble an AI app's front end instead of building it from nothing.

ai-gateway

One doorway to more than a hundred AI models from different companies: switch providers without rewriting code, see what you are spending, and cap what any one user can burn.

ai-generation-persistence

Everything an AI produces costs money and can never be reproduced exactly; this treats each one as a record worth keeping rather than output to throw away.

ai-sdk

Answers about Vercel's AI SDK read out of the copy installed in your project, so the advice matches the version you actually have rather than whatever was true last year.

aiq-deploy

Gets NVIDIA's AI-Q research server running on your own machine and confirms it is actually answering, so there is somewhere to send research questions.

aiq-research

Asks a research question of the AI-Q server you already have running, and brings back the answer with its sources still attached.

airtable-cli

Reaches into an Airtable workspace from the command line — finding the right table, reading and updating the rows in it, and changing the columns themselves.

airtable-filters

Narrowing an Airtable table down to just the rows you want, and the traps in doing it — a dropdown field, for instance, matches on a hidden ID rather than the label you see.

airtable-overview

What Airtable is and how its pieces fit together — bases, tables, fields, records, views, automations and interfaces — for anyone about to work inside someone else's.

alpha

alphafold-skill

Looks up what AlphaFold predicts a protein's shape to be, and what else is known about it, then reports that in a few lines instead of dumping the whole raw record.

android-emulator-qa

Drives an Android app running in a simulator on your computer — tapping, typing, taking screenshots — and grabs the phone's error log whenever something goes wrong.

android-performance

Recording what an Android app is really doing on a phone or emulator, so you can point at what is making it slow or eating memory instead of guessing.

appkit-interop

A Mac app written in SwiftUI sometimes needs something only Apple's older AppKit framework can do — this is how to reach across for that one piece without rewriting the screen.

attack-path-analysis

After a security scan flags something suspicious, this works out whether a real attacker could reach it, how much damage it would do, and whether it is worth reporting at all.

auth

Adds login and sign-up to a Next.js website through one of three ready-made services — Clerk, Descope or Auth0 — instead of building accounts from scratch.

beta

bgee-skill

Bgee is a database of which tissues a gene is normally active in, in healthy animals — this queries it and hands back a short readable answer rather than a wall of raw data.

bindingdb-skill

BindingDB records which small molecules stick to which proteins and how tightly; this pulls those measurements out of it without anyone hand-building the web request.

biobankjapan-phewas-skill

Takes one DNA variant and comes back with every trait and disease BioBank Japan has linked to it, whichever of the usual notations you happen to write the variant in.

biorxiv-skill

bioRxiv and medRxiv are where biology and medicine papers are posted before peer review — this looks up a preprint's details, or whether it has since appeared in a journal.

biostudies-arrayexpress-skill

Two public archives — BioStudies and ArrayExpress — hold the data behind published biology experiments, and this pulls records out of them in a form a person can actually read.

boltz-check-status

Boltz runs long molecular-prediction jobs in the cloud, and this is the afterwards part — finding a job you already started, seeing how far it got, and getting its results down.

boltz-cli-setup

Boltz's command-line program does nothing until it is installed and signed in; this clears that hurdle, including the cases where the install or the sign-in is itself what fails.

boltz-protein-design

Invents brand-new proteins, peptides, antibodies or nanobodies that stick to a target you name — molecules nobody has made yet, rather than ones picked off an existing list.

boltz-protein-screen

Ranking a set of candidate proteins or antibodies you have already got, by how firmly each one is predicted to grab hold of the target you care about.

boltz-small-molecule-adme

From a molecule's written structure alone, with no protein target in the picture, estimates how well it dissolves, how easily it crosses a membrane, and how greasy it is.

boltz-small-molecule-design

Comes up with drug-sized molecules nobody has made yet, aimed at a protein target, for when there is no existing set of compounds to search through.

boltz-small-molecule-screen

Takes a list of compounds and a protein, works out which of the compounds are likely to lock onto it, and puts the most promising ones at the top.

boltz-structure-and-binding

Predicts the 3D shape of one protein, RNA, DNA or drug molecule — alone, or locked together with another — and how firmly the pieces hold on to each other.

bootstrap

Sets a project up the careful way round — connect it to Vercel, create the database and secrets, confirm nothing is missing, and only then start the app running.

brainstorming

A rule that the assistant talks an idea through with you and gets a yes before it builds anything, with the amount of planning scaled to how big the job turns out to be.

build-chatgpt-app

Builds an app that runs inside ChatGPT — one the assistant can call on, that draws its own small interface in the conversation — working from the current official docs rather than memory.

build-run-debug

One command that stops a Mac app, rebuilds it and starts it again — plus how to read the error when the build or the launch is what went wrong.

build-zoom-bot

Planning a program that joins Zoom meetings the way a person does — sitting in, capturing what is said, and reacting while the call is still going.

build-zoom-contact-center-app

Zoom Contact Center is the customer-support side of Zoom, and this settles which platform guide a support integration needs and what has to happen at each stage of a customer conversation.

build-zoom-meeting-app

Putting a Zoom meeting inside your own app or website: which of Zoom's kits to build it on, and how people get into the call once it is there.

build-zoom-meeting-sdk-app

The hands-on half of joining or starting real Zoom meetings from your own software: getting the join to work on each platform, and working out why it fails when it does.

build-zoom-phone-integration

Zoom Phone is Zoom's actual telephone service, and this wires it into other software — so a call can be dialled from a customer record, or logged the moment it ends.

build-zoom-rest-api-app

Driving Zoom from your own code — booking meetings, fetching recordings, listing users — and getting the sign-in, the permissions and the rate limits right so it keeps working.

build-zoom-team-chat-app

Zoom has a Slack-like messaging side, and this is for building into it — a bot that posts there, messages with buttons people can press, and the code that answers when they do.

build-zoom-video-sdk-app

Building your own video-calling feature on Zoom's plumbing — your interface, your rules, and nothing that looks like a Zoom meeting to the people in the call.

build-zoom-virtual-agent

Zoom's automated support chat, dropped into your website or mobile app, including how a conversation gets handed over to a person once the bot runs out of answers.

building-ai-agent-on-cloudflare

Builds an AI assistant that lives on Cloudflare's servers, remembers each person's conversation between visits, and can go off and do work on a timer.

building-mcp-server-on-cloudflare

Puts a set of tools online, hosted on Cloudflare, that an AI assistant can reach over MCP — with a sign-in step in front of them when they need protecting.

building-native-ui

Makes a phone app built with Expo feel like it belongs on an iPhone — Apple's own controls, icons and blur effects, the small haptic taps, pages that scroll correctly around the notch — instead of a website in a phone-shaped frame.

canva-branded-presentation

Hand it an outline and a Canva slide deck comes back already wearing the company's fonts, colours and logo, ready to keep editing in Canva.

canva-translate-design

Makes a second copy of a Canva design in another language, going through every piece of text on it, and leaves the one you already had untouched.

canvas2d-data-visualization

When a chart has far too many points for the browser's usual drawing to keep up, this is how to paint it onto a canvas instead — and how a click still finds the right dot.

cbioportal-skill

cBioPortal collects cancer genome studies — tumour samples, the mutations found in them, and the matching patient records — and this pulls pieces of it back at a size you can read.

cellxgene-skill

CELLxGENE Discover is a public archive of single-cell biology datasets, and this looks up what any collection in it contains without hauling down the data itself.

chat-sdk

Write a chat bot once and it works in Slack, Teams, Discord, Telegram and the rest — the messages, buttons and menus each of them expects are handled underneath.

chatgpt-app-submission

Getting an app listed in ChatGPT means filling in a submission form — this reads the app's own code and writes most of that form for you, flagging what would fail review.

chebi-skill

ChEBI is a dictionary of the small chemicals biology cares about; this finds a compound in it and walks the family tree of what that compound is a kind of.

chembl-skill

ChEMBL records drug-like molecules, what they were measured to do against biological targets, and how known drugs work — this fetches those records a handful at a time.

choose-zoom-approach

There are many different ways to build on Zoom; this picks the smallest one that will actually do the job, and says what else you genuinely need alongside it.

chunk

Chunk is CircleCI's AI helper for build pipelines; this works out whether the task at hand belongs in its web app or at the terminal, then walks it through from either side.

circleci-builds

A CircleCI build that failed is not always broken code — it can be a flake, a stale cache, or somebody else's outage. This works out which, fixes what is genuinely yours, and refuses to hide the rest behind a retry.

circleci-cli

CircleCI runs the automated builds and tests for a project; this drives it from the command line, looking before it touches anything and rerunning only the part that actually failed.

circleci-config

Making an automated build finish faster and stop failing at random, by measuring the slowest parts first and changing only the build's own settings, never what the product does.

civic-skill

Looks up what a cancer-related gene mutation is known to mean — the treatments it points to and the evidence behind them — in the public CIViC database, and answers in a few lines rather than raw data.

clinicaltrials-skill

Finds medical studies listed on ClinicalTrials.gov, the US government's public registry of clinical trials, and reports the matches as a short readable list instead of pages of machine output.

clinvar-variation-skill

ClinVar is the public record of which gene variants are linked to disease; this pulls a variant's records out of it, or works out which of the several ID numbers in use all refer to the same variant.

cloudflare

Cloudflare sells dozens of overlapping products; this works out which one fits what you are building, then answers from Cloudflare's current documentation instead of from what the model happens to remember.

cms

Connects a Vercel site to a system where non-developers write and edit its content — Sanity, Contentful, DatoCMS, Storyblok or Builder.io — so an editor's changes reach the live site without a developer rebuilding it.

code-review

Runs the CodeRabbit reviewer over the code you have just changed, reports what it found worst-first with a suggested fix for each, and then helps you work through them.

codex-expo-run-actions

Adds a Run button to the Codex app for an Expo mobile-app project, so one press starts the app and its logs appear in the window beside your code.

comps

Works out what a listed company is worth by looking at what investors are paying for similar companies, and hands that over either as a written valuation report or as a spreadsheet you can refresh later.

conversation-intelligence

Puts Twilio's conversation analysis on a contact centre's calls and chats, so each one comes back with a summary, a read on how the customer felt and whether the agent stuck to the script — live as it happens, or once it ends.

cron-jobs

Sets up work that a Vercel site runs to a schedule — nightly, hourly, weekdays only — with the timing written correctly, the plan's limits respected and the endpoint locked so nobody else can set it off.

cuopt-user-rules

Keeps an AI assistant from guessing while it helps you with NVIDIA's cuOpt, the solver for questions like which van should take which deliveries. It has to agree the problem with you before writing any code, put the answer somewhere you cannot miss it, and never install software or take admin rights on its own.

d3-data-visualization

For a chart that an off-the-shelf charting library cannot draw — an unusual layout, a pitch or floor plan sitting behind the data, labels and callouts placed deliberately — this builds it with D3 instead.

dashboard-expert

Works on the dashboards inside Mixpanel, the product-analytics tool: reads an existing one and explains what it is showing, builds a new one, or rearranges and annotates what is already there.

dashboards-and-real-time-visualization

Design help for a screen of numbers that never stops updating: what belongs where, how often it should refresh, and how to keep it quick when the data keeps arriving.

data-visualization

The front door for anything chart-shaped: it works out what the data actually needs to show — a chart, a dashboard, a map, a diagram, a story you scroll through — and settles that design before any code is written.

dcf

Values a listed company the long way round — forecast the cash it will throw off in future, then discount it back to what it is worth today — and hands it over as an Excel workbook where the numbers are live formulas you can trace and change.

debug-zoom

When something you built on Zoom stops working, this narrows the fault to one layer, names the two to four likeliest causes and tells you how to confirm the fix — instead of you reading Zoom's whole documentation set.

debug-zoom-integration

A fixed order for hunting down a fault in a Zoom integration that used to work — log-in and app settings first, then the request itself, then the software kit, then audio and video — with the evidence to collect at each step.

deep-security-scan

Runs the same security audit over a codebase several times over, each pass independent of the others, and merges what they found — because one pass alone misses things the next one catches.

deployments-cicd

Getting a site onto Vercel and back off again: shipping a new version, promoting a tested one to live, rolling a bad one back, and setting up the automation that does all of it whenever code lands in GitHub, GitLab or Bitbucket.

dispatching-parallel-agents

When several unrelated things are broken at once, this hands each one to its own helper to work on at the same time, with each brief written so they do not trip over one another.

durable-objects

Cloudflare's Durable Objects give each chat room, each game, each booking its own single place to keep its state; this writes one for you, or reviews the one you already have against Cloudflare's current documentation.

dynamo-interconnect-check

Checks, without changing anything, whether a Dynamo serving cluster is really moving data over its fast direct links between machines and graphics cards, or has quietly dropped to a slow path that undoes the whole design.

dynamo-router-starter

Changes how a Dynamo serving cluster hands incoming requests to its workers — strict rotation, or smarter schemes that follow the load or the data already in memory — then sends a test request to prove it still answers.

efo-ontology-skill

Looks up diseases and traits in EFO, the standard vocabulary biomedical researchers name them by, and comes back with the official term and what sits underneath it in a few lines rather than a wall of data.

email

Getting a website to send real email — the welcome message, the receipt, the password reset — wired up through Resend, with the templates written like any other page and a sending address of your own.

encode-skill

ENCODE is a public library of experiments showing which stretches of DNA do what, and in which kinds of cell; this pulls one record out of it, or runs a search, and reports back in a few lines.

ensembl-skill

Asks Ensembl, the reference catalogue of genes across species, the everyday questions: what is this gene, what else sits in this stretch of DNA, what do other databases call the same thing.

env-vars

Keeps the settings and passwords a Vercel app runs on in the right place — the right value for the live site, for previews and for your own machine — and keeps the secret ones from being shipped out to the browser.

epigraphdb-skill

EpiGraphDB joins up what research has found about traits, genes and drugs; this asks it for the evidence connecting them — which drugs act on a given gene, what the literature links a trait to — and answers briefly.

eqtl-catalogue-skill

Fetches findings from the eQTL Catalogue, the collection of results on which genetic differences turn a gene's activity up or down. The public service behind it is unreliable, so its failures are reported rather than papered over.

eva-skill

The European Variation Archive is a record of the genetic differences found in humans and other species; this looks one up by its reference number, or lists which species the archive covers.

evaluate-plugin

Grades a whole Codex plugin you have built, every skill inside it, and puts the findings in order so you know which one to deal with before the rest.

evaluate-skill

Reviews one skill you have written — is the description doing its job, are the linked references earning their place, what does it actually cost in tokens — and names the first thing to fix.

executing-plans

For when the plan is already written and all that is left is doing it: one task at a time, each one checked before the next begins, and a stop-and-ask the moment something does not match.

expo-api-routes

Adds a piece of server code to an Expo mobile app for the work that must not happen on the phone itself, and puts it online on Expo's own hosting.

expo-cicd-workflows

Writes the file that tells Expo's build service what to do on its own — build on every push, run the tests, send a release to the stores — and checks it against what the service accepts today.

expo-deployment

The last mile for a phone app built with Expo: turning it into the file Apple and Google will accept, and getting it into their stores — or onto the web.

expo-dev-client

Builds a test copy of your own app and gets it onto a phone, a simulator or TestFlight — the version you need once the project goes beyond what Expo's ready-made preview app can run.

expo-module

Lets an Expo app reach the things only the phone's own platform can do, by writing the Swift and Kotlin side and the JavaScript that calls into it — scaffolded so that it builds the first time.

expo-tailwind-setup

Tailwind is the shorthand a lot of web developers style pages with; this is the fiddly setup that makes the same shorthand work in an Expo phone app, on iPhone, Android and the web at once.

expo-ui-jetpack-compose

Builds an Android screen out of Google's own interface pieces from inside an Expo app, so the result moves and looks like an Android app rather than a cross-platform imitation of one.

expo-ui-swift-ui

On iPhone, Apple's own interface pieces can be used directly from an Expo app's JavaScript — and an existing React Native view can be dropped into the middle of one of those Apple-built layouts.

figma-code-connect

Ties a component in a Figma design to the component that actually implements it in the codebase, so that anyone opening that design in Figma is shown the real code for it.

figma-create-new-file

Opens an empty Figma file to start in — a design, a whiteboard or a slide deck — and picks which team it should belong to when you are in more than one.

figma-generate-design

Builds a whole screen in Figma out of the buttons, colours and text styles the file already has, instead of drawing fresh boxes that only look like them.

figma-generate-diagram

Draws a diagram on a FigJam board from a written description of it — a flowchart, the back-and-forth between two systems, a data model, a project schedule — and it stays editable in Figma afterwards.

figma-generate-library

Takes the colours, spacing, type and components a codebase already uses and reproduces them in Figma, so designers are working from the same set of parts the engineers are.

figma-implement-motion

A designer animated something in Figma; this turns that animation into real code in the app, keeping the exact timing and movement the designer chose rather than approximating it.

figma-swiftui

Translates between a Figma design and the code of an iPhone or iPad app, in whichever direction is needed — design turned into screens, or screens already built pushed back into the design file.

figma-use

Ground rules for letting an assistant build and edit a Figma file by running code inside it, collected from the mistakes that quietly break the file on a first attempt.

figma-use-figjam

FigJam is Figma's online whiteboard, and this is what it takes to build one by code — sticky notes, sections, arrows, shapes and tables — including the rules that differ from an ordinary Figma file.

figma-use-motion

The animation side of working in Figma: what an assistant needs in order to make things move inside the design file itself, instead of writing the animation in code.

figma-use-slides

Builds a presentation in Figma Slides the way a designer would — settle the look first, plan every slide, then build — instead of pouring the content into a generic template.

finding-discovery

The first pass of a security review: goes through the code for anything that could plausibly be a vulnerability and writes each one down with where it sits and why it looks wrong, leaving the question of whether it is real to a later stage.

finishing-a-development-branch

The end-of-task routine for a piece of coding work: run the tests, then ask whether it should be merged, opened as a pull request, or left alone — that call stays with the person, never the assistant.

finngen-phewas-skill

FinnGen is a large Finnish study that links people's genes to their medical records, and this pulls one variant's slice of it back as a short readable list instead of a wall of data.

fix-finding

Takes one reported security hole and closes it with the smallest change that genuinely works — or, when the code turns out to have been safe all along, proves that and changes nothing.

frontend-app-builder

Designs the screen as a picture first, gets that approved, then builds the real thing in code and keeps fixing it until what the browser shows matches the picture.

frontend-testing-debugging

Opens the app in a real browser, clicks through the part that is misbehaving, fixes it, and then shows screenshots and error logs as proof rather than simply claiming it works.

game-playtest

Plays a browser game the way a player would — boots it, presses the buttons, moves between screens — and reports what is broken, worst first, with screenshots and steps to repeat it.

game-studio

The first conversation about a browser game, before any code is written: what kind of game it is, which toolkit to build it with, and who picks up each piece — the art, the interface, the testing.

game-ui-frontend

Menus, scoreboards and on-screen readouts for a browser game, designed to look like they belong in that game's world and to stay out of the way of what the player is actually doing.

gantt-chart-visualization

Gantt charts — the bars that show when each piece of a project starts and finishes — done properly, starting with whether a Gantt chart is even the right picture for the data.

geist

Geist is the typeface Vercel uses across its own products, and this covers getting it into a website plus the house rules for which of its three styles belongs where.

geistdocs

A ready-made documentation website from Vercel that you fill with content instead of building from scratch — search, navigation, translations and deployment come already wired up.

genebass-gene-burden-skill

Genebass pools the rare mutations found in a single gene and asks what they turn out to be associated with; this fetches that answer for one gene and reports it as a short list.

geospatial-and-cartographic-visualization

Maps for data, treated as a decision rather than a default: first whether a map genuinely helps at all, then what kind of map it should be and what it sits on top of.

gnomad-graphql-skill

gnomAD is the public reference for how common each human genetic variant is, and this asks it one narrow question at a time and brings back a short answer rather than a data dump.

google-docs

Writing and editing in Google Docs, with one rule doing most of the work: given an existing document to follow, it copies that document and works inside it rather than rebuilding something that merely looks similar.

google-drive

The place to start for anything in Google Drive: it works out which file you mean and handles it as a file, then hands over to a specialist the moment the job becomes editing what is inside.

google-drive-comments

Commenting on Google files — leaving notes, answering them, closing them off — written so that every note points at something the reader can actually find on the page.

google-sheets

Spreadsheet work that stays pinned to the actual workbook, the actual rows and the actual formulas in play, instead of guessing at tab names and writing over the wrong thing.

google-slides

Point it at a company template or last quarter's deck and it works inside that design — filling it in, editing it, tidying it up — rather than starting a fresh one that looks nothing like it.

grammar-of-graphics-and-declarative-visualization

Most charts do not need drawing code written by hand — you describe the chart and a library draws it. This covers the three libraries that work that way and which one to reach for.

gsap

GSAP is the library a great deal of web animation is built with, and this is the working reference for it: how to move something, how to sequence several movements, and how to keep it smooth.

gtex-eqtl-skill

GTEx records which genes a genetic variant turns up or down, and in which tissue of the body; hand this one variant and it returns that list as plain data.

gwas-catalog-skill

The GWAS Catalog is the public record of published studies linking genes to traits and diseases, and this queries it for a short answer rather than the raw feed.

hmdb-skill

HMDB is the reference database of the chemicals found in the human body, and this searches it for a compound, protein, disease or pathway and reports back a short answer.

human-protein-atlas-skill

The Human Protein Atlas is the public map of where in the body every human protein turns up; this pulls an entry out of it and boils it down to a few readable lines.

hyperframes

A way of building a video by writing it out as a web page — scenes, captions, voiceover and music placed on a timeline and animated into view.

hyperframes-cli

The terminal side of HyperFrames — everything between an empty folder and a finished video file runs through a single command.

hyperframes-registry

A shared library of ready-made video parts you can drop into a HyperFrames video and wire into its timeline.

improve-skill

Takes a report on what is wrong with one of your skill files and turns it into a concrete plan for rewriting it.

investigation-mode

An app has stopped responding and nobody knows why — this works down a fixed list of places to look and says what it found at each one instead of going quiet on you.

ios-app-intents

Picks out the handful of things an iPhone app does that people would want from Siri, Shortcuts or a widget, and wires up just those rather than the whole app.

ios-debugger-agent

Runs your iPhone app on the simulated phone that lives on your Mac, then taps through it and reads its logs to work out what is going wrong.

ios-ettrace-performance

Measures where the time actually goes when an iPhone app does one particular thing — starting up, say, or opening a screen — and names the code that is eating it.

ios-memgraph-leaks

Proves that an iPhone app really is hanging on to memory it should have let go of, shows which piece of code is still clinging to it, and makes the smallest change that lets go.

ios-simulator-browser

Puts a running iPhone simulator inside Codex's own browser window, so you can watch an app and click around in it without switching over to Xcode.

ipd-skill

Answers questions about HLA types — the immune-system markers that decide whether a donor and a patient match — out of the official public catalogue of them.

json-render

Fixes a chat app that is showing raw message data instead of a readable reply.

liquid-glass

Brings a Mac app onto Apple's new frosted-glass look, mostly by deleting the custom backgrounds that were fighting it rather than by adding more effects.

locus-to-gene-mapper-skill

Genetic studies point at a spot on the genome linked to a disease but not at the gene responsible; this weighs the evidence and hands back a ranked shortlist of the likely ones.

magicpath

MagicPath is a drawing board where a team's UI components live; this moves them into your codebase, or takes screens you already built the other way onto the board.

marketplace

When an app needs a shop, a payment button or email sending, this picks a ready-made service and sets it up for real — so nothing gets built on a stand-in you will throw away later.

metabolights-skill

Finds studies in MetaboLights, the public archive of experiments that measured the chemical make-up of a blood, tissue or cell sample, and says what each one covered.

metric-pack-designer

Adds your team's own grading criteria to the skill checker, as a small script it runs alongside its own built-in ones.

mgnify-skill

Microbiome research catalogues which microbes live where — in a gut, in soil, in seawater. This reaches into MGnify, the public record of those studies, and reports a study, sample or habitat back in short form.

micro

micro is Vercel's stripped-down way of writing a web server that does exactly one thing, in a handful of lines with almost no boilerplate.

minimal-plugin-skill

minimal-skill

mixpanel-auth

Gets you signed in to Mixpanel and keeps you there — switching between accounts and projects, checking the login still works, and setting up a token for the places where nobody can click a browser.

mixpanel-headless-setup

The one-time preparation before you can analyse Mixpanel data from Python: get the packages onto the machine, then check that the login actually works.

mixpanelyst

Answers questions about how people actually use your product — where they drop off, who comes back, what they do next — by querying Mixpanel from Python.

native-data-fetching

How a phone app built with Expo or React Native should talk to a server, and what to do when that goes wrong.

ncbi-blast-skill

Takes a stretch of DNA or protein and asks NCBI which known sequences it resembles, then reports the closest matches rather than the full alignment dump.

ncbi-clinicaltables-skill

Type part of a gene's name and get the matching human genes straight back — the quick search-box kind of lookup, not a full database record.

ncbi-datasets-skill

Pulls the facts NCBI holds about a genome — which organism, which assembly, where it sits in the tree of life — one record at a time.

ncbi-entrez-skill

The one door into NCBI's whole family of databases — papers on PubMed, genes, proteins, DNA sequences — with every paper reference coming back as a link you can click.

ncbi-pmc-skill

Tells you whether a given research paper is free to download from PubMed Central.

ncc

Squashes a Node.js program and every library it leans on into one file, so what you ship is a single file rather than a folder of thousands.

nemoclaw-user-get-started

Walks you from nothing to a working NemoClaw install — NVIDIA's walled-off box for running an AI agent on your own machine — with its first agent already answering.

next-forge

A ready-made starting point for a paid web product — the marketing site, the signed-in app and the back end all come pre-built, with logins, payments, database and email already connected.

nextjs

Next.js is one of the common ways to build a website; this is the collected advice on doing it right — which approach to pick for a given job, and what is really causing the errors it throws most often.

ngs-amplicon-microbiome

Works out which bacteria, fungi or other microbes are present in a sample from its DNA sequencing files, and how the mix compares across samples.

ngs-analysis-router

The front desk for DNA sequencing work: it looks at the files you already have, works out which kind of analysis they call for, and asks only the questions the files cannot answer.

ngs-atacseq-peaks-qc

Finds which stretches of DNA are sitting open in a cell — the places where genes can actually be switched on — from an ATAC-seq run's raw sequencing files, and flags any sample too poor in quality to trust.

ngs-bcl-to-fastq

Turns the raw output of an Illumina sequencing machine into one readable data file per sample, and checks the samples really were told apart correctly before anyone analyses them.

ngs-bulk-rnaseq

Sorts a gene-activity experiment into its right next step — measuring how active each gene was from the raw sequencing files, or comparing those measurements between groups of samples.

ngs-bulk-rnaseq-counts-qc

Turns raw sequencing reads into a table of how active each gene was in each sample, with the quality checks that say whether those numbers can be trusted.

ngs-bulk-rnaseq-differential-expression

Takes a table of gene activity and says which genes genuinely differ between two groups of samples — treated against untreated, say — and which comparisons have too few samples to answer honestly.

ngs-chip-cutrun-peaks-qc

Maps where a particular protein sits along the genome in a ChIP-seq, CUT&RUN or CUT&Tag experiment, and checks the signal is strong enough before anyone reads meaning into it.

ngs-dna-germline-variants

Finds the DNA differences a person was born with — for one person, a family or a whole cohort — and checks coverage and sample identity before that list is believed.

ngs-dna-somatic-variants

The DNA mutations a tumour picked up during life rather than inherited — read against matched healthy tissue where there is one, with the confident calls sorted from the likely artefacts.

ngs-dna-umi-panel-variants

Picks out mutations carried by only a tiny fraction of the DNA in a sample, using the molecular tags that show how many separate original molecules really support each call.

ngs-dna-variant-calling

One request, three very different analyses: this works out whether the DNA changes you are after are inherited, picked up by a tumour, or the rare kind a barcoded panel hunts, and hands the job to the matching skill.

ngs-epigenomics-peaks

Tells apart the two kinds of experiment that map things onto the genome — the one that finds open stretches of DNA and the ones that track a specific protein — and hands the job to the workflow built for it, after one shared check of the raw reads.

ngs-fastq-qc

The first look at raw sequencing data — how good the reads are, whether leftover lab sequence needs trimming off, and whether anything looks wrong enough to stop — with the original files always left untouched.

ngs-runtime-env

Checks whether a machine already has the sequencing software and reference data a pipeline needs, and writes out a reviewable plan for whatever is missing instead of quietly downloading it.

ngs-scrna-seq

Single-cell sequencing measures gene activity one cell at a time; this covers the front half of that work — raw files in, a per-cell table of gene counts out — and then passes it on for analysis.

ngs-shotgun-metagenomics

Takes DNA sequenced from a whole community — a gut, a soil sample, a water sample — and works out which organisms are in it and what chemistry they are capable of, stripping the host's own DNA out first if asked.

node-link-and-diagram-layout

How to arrange a diagram of boxes joined by lines so that it can actually be read — which arrangement suits which kind of diagram, and how to stop the lines crossing, overlapping or disappearing behind the boxes.

notion-knowledge-capture

Writes up what a conversation settled as a proper page in your Notion workspace — a how-to, an FAQ, a decision record — filed in the right place and linked to whatever it relates to.

notion-meeting-intelligence

Prepares a meeting before it happens: digs the relevant pages out of Notion, picks a format that suits the kind of meeting, and drafts the agenda with an owner and a time for each item.

notion-research-documentation

Answers a question by reading across your Notion workspace and writing up what it found as a brief, a comparison or a full report, with each claim linked back to the page it came from.

notion-spec-to-implementation

Reads a product spec sitting in Notion and turns it into a build plan plus a set of tasks small enough to finish in a day or two, each one linked back to the part of the spec it came from.

observability

Seeing what a live site on Vercel is actually doing — what it logged, what broke, how fast the pages felt to real visitors — and getting that stream out to whatever tool you watch it in.

omniverse-cad-to-simready

Turns an engineering CAD model or a robot description file into a 3D asset a physics simulator can use, with the materials and physical properties filled in, checked, and bundled ready to drop into a scene.

omniverse-realtime-viewer

How to build an app that shows a live 3D scene — in a browser, on the desktop, or streamed from a server — with NVIDIA's own renderer drawing it rather than a browser 3D library.

omniverse-usd-performance-tuning

Tracks down what is making a 3D scene slow to load, heavy on memory or jerky to move around, then measures it again afterwards so the improvement is proven rather than claimed.

openai-ads-conversions-setup

Adds the tracking that tells OpenAI Ads when a visitor actually signed up or bought something — from the browser, from your server, or both without the same purchase being counted twice.

openai-api-troubleshooting

Works out why a call to OpenAI failed — no route out to the internet, a bad key, no credit left, or simply too many requests too fast — and says what to do about that particular cause.

openai-platform-api-key

The credential step before anything is built against OpenAI: it checks whether a usable key already exists, asks before making a new one, and writes it to a file you confirm without ever showing the key itself.

opentargets-skill

Looks a gene, a disease or a drug up in Open Targets — the public database that connects the three — and answers with a short summary rather than raw data, plus a grid of which evidence sources back each link when you ask for one.

packaging-notarization

Before a Mac app can be handed to anybody outside, Apple has to sign off on it — this checks whether yours would pass, and names the one thing stopping it if it would not.

payments

Taking money in an app hosted on Vercel: one-off payments, subscriptions a customer can manage themselves, and the plumbing that tells your app when a payment really went through.

pharmgkb-skill

How a person's genes change the way a drug works for them: this pulls PharmGKB's public data on gene-drug pairs, variants and dosing guidance and answers in a few lines.

phaser-2d-game

Building a 2D browser game with Phaser, keeping the rules of the game apart from the code that draws it so the visuals can be redone without breaking the gameplay.

physical-ai-infrastructure-setup-and-resilient-scaling

Sets up the GPU cluster, model-serving endpoints and job scheduler that NVIDIA's synthetic-data workflows for robotics run on — on your own machines or in Azure — and gets them back up when one of those layers falls over.

physical-ai-neural-reconstruction

NuRec turns camera and LiDAR recordings from a car or a robot into a 3D scene you can re-render from any viewpoint; this is the front page that says which of its five skills does the part you need, and in what order.

plan-zoom-integration

Turns an idea for something built on Zoom into a staged build plan — the permissions it will need, the things likely to bite you, and the smallest version worth building first.

plan-zoom-product

Zoom gives you about eight different ways to build on it; this picks the one that fits what you are actually trying to do, and says why the others are worse for it.

plugin-creator

Sets up an empty new Codex plugin — the folder layout and a starter settings file with every blank marked — and, if you want, lists it so Codex shows it alongside your others.

plugin-eval

The way in when you do not know which check you need: point it at a skill or a plugin on your machine and it picks the right one, explains the score in words, and says what to run next.

posthog

Asks your product's own usage data the questions you would otherwise click through PostHog for — how many people finished signing up, where the rest dropped out — and can switch a feature on for a test group while it is in there.

pride-skill

PRIDE is the public archive where labs deposit their protein mass-spectrometry experiments; this searches it by keyword and reports back what a given project contains, in a few lines rather than a data dump.

probe-sdk

Before someone joins a Zoom call in your app, this tests their camera, microphone, speakers and connection, so a problem turns up while there is still time to deal with it rather than in front of everyone.

proteomexchange-skill

Searches ProteomeXchange, the index that ties the world's public protein-measurement datasets together, down to a single spectrum or protein, and answers with a short readout.

pubchem-pug-skill

PubChem is the public catalogue of chemical compounds; this pulls out just the part you asked for — a compound's formula and weight, a plain description of what it is, or how it performed in a lab assay.

quickgo-skill

The Gene Ontology is the standard vocabulary for what a gene or protein actually does; this looks a term up, shows where it sits in that hierarchy, and finds the proteins labelled with it.

rcsb-pdb-skill

The Protein Data Bank holds the experimentally solved 3D structures of proteins and other large molecules; this searches it, says what an entry actually is, and can fetch its sequence file.

react-and-nextjs-data-visualization

Charts and diagrams that sit inside a website built with React or Next.js, put together so the page still loads quickly and nothing flickers or breaks on the way in.

react-best-practices

Vercel's list of 64 rules for making a React or Next.js site load and respond faster, written to be applied while the code is being written or reviewed rather than after it is slow.

react-three-fiber-game

Build a 3D game or 3D scene that runs in a web browser and lives inside an app you already built with React, so the 3D part and the ordinary page share the same data and controls.

reactome-skill

Looks things up in Reactome, a free public database of how proteins and genes interact inside a cell, and answers with a short readable summary instead of a raw data dump.

receiving-code-review

How to answer someone's comments on your code: check whether they are actually right before you change anything, say why when they are not, and drop the flattery.

remotion-best-practices

Remotion turns code into video; this is the front desk that sends you to the right set of Remotion notes for what you are doing, from making a video to adding captions to rendering the finished file.

reports-pdfs-and-slide-automation

Building the finished document - a PDF report, a slide deck, a Word file - where every chart in it is a properly exported piece of artwork that can be regenerated when the numbers change, rather than a screenshot pasted in.

requesting-code-review

Hands your finished changes to a fresh assistant that has not watched you work, so what comes back judges the code as it now stands rather than the story of how you got there.

research-router-skill

Takes a broad biology or medicine question, works out which specialist databases can answer which part of it, collects from each, and writes back a single answer that says what the evidence supports and where it disagrees with itself.

rhea-skill

Rhea is a free catalogue of the chemical reactions that take place inside living cells; this pulls one reaction out of it, or searches it by substance, and gives the result back as a short readable summary.

rivet-sdk

Builds a program that plugs into Zoom using Rivet, Zoom's own toolkit, so signing in, calling Zoom and receiving its notifications are handled for you instead of written from scratch.

rnacentral-skill

Fetches an entry from RNAcentral, the public catalogue of RNA sequences collected from every species, together with the other databases that describe the same molecule.

routing-middleware

A small piece of code that inspects every visit to a site hosted on Vercel before anything is sent back, and can send the visitor elsewhere, quietly serve different content, or shape the page around who and where they are.

runtime-cache

A short-term memory a Vercel app keeps in each part of the world it runs in, so something worked out once is not worked out again for the next visitor, and can be wiped in bulk once the underlying data changes.

sandbox-sdk

A sealed box running on Cloudflare's network where code you have reason not to trust - code an AI just wrote, or code a stranger pasted in - can run without touching anything else.

satori

Builds the little preview picture that appears when someone pastes a link to your site into a chat or a social post, drawing it from ordinary web-page markup instead of a design tool.

scribe

Turns recorded Zoom audio and video into written text using Zoom's own transcription service, and helps you pick between the quick answer and the slower route built for long or bulk files.

scrna-seq-qc

Cleaning up a single-cell gene-activity experiment before anyone draws conclusions from it: throw out the cells that were damaged or were secretly two cells stuck together, work out what type each survivor is, and keep a record of every judgement call made along the way.

scrollytelling-and-parallax-data-visualization

Stories where the picture changes as the reader scrolls - a map that travels, a chart that builds itself, a timeline that advances - planned scene by scene, and still readable for someone who has animation switched off.

security-diff-scan

Looks only at what a change actually touched - a pull request, one commit, or edits not yet committed - for security holes it introduces, and deliberately stops there instead of auditing the whole project.

security-scan

Reads an entire codebase from top to bottom hunting for security weaknesses, then writes up what it found with the evidence behind each one.

sentry

Pulls up the crashes and errors your live app has been reporting into Sentry, and the detail behind any one of them, without changing anything there.

setup-zoom-oauth

Getting an app permission to act on somebody's Zoom account, and keeping that permission alive afterwards: which kind of sign-in to ask for, how little access to request, and where the renewal usually breaks.

setup-zoom-webhooks

Zoom can call your own server the instant something happens in a meeting; this is how to take those calls safely - proving each one really came from Zoom, and not acting twice when the same one arrives twice.

setup-zoom-websockets

Holds one connection open to Zoom and takes its events down that line instead of having Zoom call your server, which is the option when your server cannot be reached from outside - and makes the reconnecting your job.

shadcn

shadcn/ui hands you ready-made interface pieces - buttons, dialogs, tables - as source code copied into your own project rather than as a package you install. This is about getting them in, giving them one consistent look, and assembling real screens out of them.

shopify-admin

Writing the request a program sends to Shopify to read or change a shop's products, orders and settings, checked against Shopify's own current documentation before anyone runs it.

shopify-app-store-review

Marks an app you built for Shopify's app store against the store's published rules and says, rule by rule, what passes, what fails and what a person will have to judge - before you submit it for the real review.

shopify-custom-data

Adds fields Shopify does not ship with - a care guide attached to a product, an author record of your own design - by declaring them up front, then writing values into them and reading them back.

shopify-customer

Writing the request that lets a shopper who is signed in to a Shopify store see and manage their own things - past orders, saved addresses, payment cards - and nobody else's.

shopify-dev

Searches the whole of Shopify's developer documentation, for the questions that straddle several parts of Shopify at once and have no more specific place to go.

shopify-functions

Shopify decides for itself how a discount is worked out and which delivery and payment options a shopper sees; this replaces those with rules you write yourself.

shopify-hydrogen

Hydrogen is Shopify's framework for shops that want to build their own website instead of using a ready-made theme, and this is the worked-example book for it.

shopify-liquid

The look of a Shopify shop — its banners, product grids and the sections merchants can edit themselves — comes from theme files, and this writes them.

shopify-onboarding-dev

The very first step for someone who wants to build something for Shopify: it gets the tools onto your machine and into your editor, then points you at whichever skill fits what you are making.

shopify-onboarding-merchant

For shop owners rather than developers — it opens a new Shopify store or connects one you already have, then lets you run the shop from your AI assistant.

shopify-partner

If you sell apps or themes through Shopify, your Partner dashboard holds the numbers — installs, payouts, referral commissions — and this gets them out where you can use them.

shopify-payments-apps

Payment providers plug their own way of paying into Shopify's checkout, and every step of the money moving is a call back to Shopify; this writes those calls.

shopify-polaris-admin-extensions

Puts your app's own buttons, cards and print options directly onto the pages where a merchant runs their shop, so they never have to leave for your app.

shopify-polaris-app-home

An app that a shop installs gets its own screens inside Shopify, and this builds them out of parts that already look and behave like the rest of Shopify.

shopify-polaris-checkout-extensions

Shopify's checkout pages are closed to outsiders, but an app is allowed to add its own content at set spots, and this builds those pieces.

shopify-polaris-customer-account-extensions

Once someone has bought something they can log in to follow their order, and this adds your app's own content to those pages.

shopify-pos-ui

Shopify POS is the till app that staff use in a physical shop, and this adds your own buttons, screens and receipt text to it.

shopify-storefront-graphql

When a shop's website is built from scratch instead of from a Shopify theme, every product and basket has to be fetched by hand — this writes those requests.

shopify-use-shopify-cli

Most Shopify work is actually run by typing commands into a terminal; this picks the right one for what you are doing and sorts it out when it fails.

sign-in-with-vercel

Lets people log into your app with the Vercel account they already have, so you never store a password or run a sign-up form of your own.

signing-entitlements

A Mac refuses to open an app, or it quits the moment it starts: this works out whether the signature, the permissions or Apple's security check is to blame, and what the shortest fix is.

sprite-pipeline

A game character's animation is a row of small drawings, and this makes the whole row from one drawing you have already approved, then trims every frame to the same size so the character does not wobble.

start

The way in to anything Zoom: it works out what you are really building — video inside your own app, a bot that sits in meetings, a login flow — and sends you to the instructions for that.

statistical-and-uncertainty-visualization

Averages and smooth lines can hide how shaky the numbers underneath really are; this picks the chart that shows the spread, the uncertainty and the gaps honestly.

string-skill

STRING is a public database of which proteins work with which; this asks it about one protein and answers in a few readable lines instead of raw data.

stripe-best-practices

Stripe offers several different ways to take a payment and they are easy to mix up; this picks the one that fits what you are building and keeps you off the ones Stripe has retired.

subagent-driven-development

Hands a written plan to a series of fresh assistants, one task each, reviews every task the moment it lands, and runs the whole plan to the end without stopping to ask you.

supabase

Supabase gives an app a ready-made database, logins and file storage; this is how to work on one safely — checking what changed recently rather than trusting memory, and catching the settings that quietly leave data open.

supabase-postgres-best-practices

A rulebook for keeping a Postgres database fast — the missing index, the query that reads every row, the connection settings that give way under load — ordered by how much each one costs you.

swiftpm-macos

Builds, runs and tests a Swift project that has no Xcode project to open, and says plainly which of the build, the dependencies, the linking or the tests is what broke.

swiftui-liquid-glass

iOS 26 gave Apple's buttons and panels a frosted-glass look; this puts it into an app properly — grouped, only where it belongs, and with something sensible on phones too old to show it.

swiftui-patterns

Mac apps have furniture that phone apps do not — several windows, a menu bar, a settings pane, a sidebar — and this lays one out around that from the start.

swiftui-performance-audit

An app that stutters when you scroll almost always has a specific cause in the code; this hunts for it there first and only asks you to record a performance trace when reading the code cannot settle it.

swiftui-ui-patterns

Builds the screens of an iPhone app — the tabs, the way one screen pushes to the next, the panels that slide up — while keeping to whatever the project already does.

swiftui-view-refactor

One screen's file has grown until nobody wants to touch it; this breaks it into small named pieces and moves the decisions out of the layout code.

swr

SWR is a small React library that fetches data, keeps a copy and quietly refreshes it; this is how to use it instead of writing loading flags and caches by hand.

systematic-debugging

A rule rather than a tool: work out what is actually causing a bug before changing a line, however obvious the fix looks and however urgent it feels.

tearsheet

A one-page factual profile of a listed company — what it does, what its numbers say, what could go wrong — with a source behind every figure and no opinion on whether to buy it.

telemetry

Adds a handful of well-chosen log lines to a Mac app — window opened, menu item picked, sync finished — then checks each one really shows up when you use the app.

template

Pick a ready-made Adobe Express design — a flyer, a poster, a social post, a CV — and this does the editing for you and hands back something you can share or download.

temporal-developer

Temporal keeps a long-running job alive through crashes and restarts by replaying what already happened; this builds and debugs those jobs in any of the seven languages it supports.

test-driven-development

Write the test, watch it fail, then write just enough code to make it pass — and delete anything that was written before the test existed.

test-triage

A failing test is not always a broken product; this works out whether it is the build, the test, a crash, a flake or the setup, and reruns only the part that proves it.

testing-data-visualizations

Charts go wrong in ways an ordinary test never notices — numbers wrongly added up, a hover that stopped working, a layout that collapses on a phone — so this decides what to test, at what level, and what to leave alone.

threat-model

Writes down how a codebase could actually be attacked — what it is made of, where it trusts the outside world, what an attacker could reach — or reuses the one on file when nothing has changed.

three-webgl-game

Making a 3D game that runs in a web browser, with hands-on control of the scene, the camera and the frame loop instead of a framework doing it for you.

threejs-data-visualization

Some data will not fit in a flat chart — millions of points, real depth, a map you can fly around — and this draws it with the graphics card, while saying plainly when a plain 2D chart would serve better.

tpmi-phewas-skill

Takes one spot in the genome and reports which traits and diseases it has been associated with in the TPMI database, accepting whichever of the usual naming formats you happen to have.

track-findings

Files a confirmed security problem as a ticket in Linear, Jira or GitHub — or as a private draft advisory — after showing you exactly what it will write and checking nobody filed it already.

triage-finding

A pile of security alerts from scanners and old tickets, most of which will not apply — this checks each one against the actual code, says whether it is real, and ranks the ones that are.

turbopack

Turbopack is the tool Next.js 16 uses to compile a site's code — this covers configuring it and the places it behaves differently from the older webpack setup it replaced.

turborepo

When one repository holds many projects, Turborepo keeps their builds fast by remembering what it already built and skipping anything that has not changed.

twilio-account-setup

Twilio is the service apps use to send texts and make phone calls; this is the one-time account setup that every other Twilio skill assumes is already done.

twilio-agent-augmentation-architect

A planning conversation for a call centre that wants AI listening in alongside its human staff: it settles what the AI should actually do for them, and what to build it from.

twilio-agent-connect

Twilio Agent Connect is the middle layer between an AI assistant and a real phone number — the assistant does the thinking, this handles the call, the texts and the customer's history.

twilio-ai-agent-architect

Before any code is written, this asks the questions that settle what kind of AI phone or chat assistant is really needed, and names the Twilio parts to build it from.

twilio-call-recordings

Twilio offers several different ways to record a phone call and the obvious-looking one is usually wrong; this is how to pick and wire up the one that fits.

twilio-cli-reference

Twilio's command-line tool handles the account chores — buying a number, sending a test text, pointing a number at your app — with a typed command instead of written code.

twilio-compliance-onboarding

Messages get blocked and calls get labelled as spam when the paperwork was never filed — this is what the carriers and platforms require up front, and how long each approval takes.

twilio-compliance-traffic

The law limits who you may text or call, at what hours, and what you may record or keep — this is how to run live traffic without falling foul of those rules.

twilio-conference-calls

Putting a caller into a conference room instead of connecting two phones directly is what later allows a hold, a handover to a colleague, or a supervisor listening in.

twilio-content-template-builder

Write a message once with blanks left in it — a name, a date — get it approved where WhatsApp demands that, and send it out filled in over any channel.

twilio-conversation-orchestrator

A customer who phones on Monday and texts on Tuesday normally arrives as two strangers; this gathers everything into one running thread per person, without your code stitching it together.

twilio-conversations-classic-api

Twilio's older system for a shared thread several people can talk in — one on WhatsApp, one on text, your staff on a web page — with the whole history kept in one place.

twilio-customer-memory

Remembers each customer between conversations — what they are like, what happened last time — and hands the relevant pieces back when they next get in touch.

twilio-customer-support-architect

Someone asks for a support line: this works out whether that means a simple phone menu or a full contact centre with call routing and agent desks, and what to build it from.

twilio-debugging-observability

A text never arrived and nobody knows why: this is how to find out what actually went wrong, and how to be told the next time before a customer complains.

twilio-email-deliverability-advisor

Why email sent through Twilio's own email service lands in spam instead of the inbox, and what to change so it stops — SendGrid mail has separate advice of its own.

twilio-email-send

Sends email from your own code through Twilio — one message, or up to ten thousand at once, each one carrying the recipient's own details.

twilio-enterprise-knowledge

An AI answering your customers knows nothing about your warranty terms or your refund policy; this puts your own documents somewhere it can look them up and answer from them.

twilio-iam-auth-setup

The keys that let your code act on your Twilio account — which kind to use while you are still experimenting, which once real customers depend on it, and how to replace one that leaks.

twilio-identity-verification-advisor

Deciding how a new user proves they are really themselves — a code by text, a push, an authenticator app — and how much checking the business actually needs.

twilio-isv-sms-best-practices

If you sell software that sends texts on your customers' behalf, each of those customers needs their own registration, their own numbers and their own bill — this is how to lay that out.

twilio-lookup-phone-intelligence

Checks what is really behind a phone number before you text or call it — whether it exists, whether it is a mobile, and whether it looks like fraud.

twilio-marketing-promotions-advisor

Sending an offer to a list of customers comes down to which channel, whose permission you already have, and what the law allows — this settles all three before a single message goes out.

twilio-messaging-channel-advisor

People ask for a text message out of habit; this weighs where the recipients live, what is being sent and what it costs, then says whether plain text is really the right choice.

twilio-messaging-overview

The starting map of Twilio messaging: what each channel is good for, from a plain text to WhatsApp, and the order to do things in from a first test message to live traffic.

twilio-messaging-services

Rather than send every message from one phone number, you hand Twilio a pool of them to choose from — which is also where scheduling, opt-out handling and fraud protection get switched on.

twilio-messaging-webhooks

When somebody texts your number back, Twilio calls your server — this is how to catch that, answer it, and follow whether your own messages ever arrived.

twilio-notifications-alerts-advisor

An order update, an appointment reminder, an outage alert: how urgent each one is decides how it should be sent, whether it needs a second attempt elsewhere, and whether a reply matters.

twilio-numbers-senders

The number your messages come from is the first thing to pick and the most expensive to change later — an ordinary number, a toll-free one, a short code, or a name instead of a number.

twilio-organizations-setup

Once a company has several Twilio accounts and several people using them, they all sit under one organisation — this is where staff, roles and company-wide logins are managed.

twilio-rcs-messaging

RCS messages arrive in the phone's ordinary texting app, but carry your logo, your name and buttons people can tap — this is what it takes to get approved to send them.

twilio-regulatory-compliance-bundles

Many countries will not let you hold a local phone number until you have proved who you are and where you are based; this is the paperwork that proves it and how to get it accepted.

twilio-reliability-patterns

Send enough messages fast enough and Twilio starts turning them away; this is how to back off, retry sensibly, and keep a large send from collapsing under its own volume.

twilio-security-api-auth

Twilio accepts four different kinds of credential, and the easiest one is the dangerous one — this is which to reach for in a test, on a server, and in code running inside someone's browser.

twilio-security-compliance-hipaa

Putting American patients' health information through Twilio is allowed only under a signed agreement, on accounts flagged for it, and with the subset of products it covers — this is that sequence.

twilio-security-hardening

The ways a Twilio account actually gets abused — a leaked key, a faked incoming request, traffic to countries you never serve — and what to shut off against each of them.

twilio-send-message

One instruction sends the message, whether it goes out as a text, a picture, a WhatsApp message or a tappable card; the address you send to is what decides.

twilio-sendgrid-account-setup

SendGrid is the email half of Twilio and signs in with a key of its own; this is the first-time setup, including proving you own the domain your mail claims to come from.

twilio-sendgrid-deliverability-advisor

Mail sent through SendGrid that lands in the spam folder or gets blocked before it arrives always has a reason; this walks the usual causes in order and says which one to repair first.

twilio-sendgrid-email-send

The actual sending call for SendGrid mail: one receipt, or a whole mailing list with each copy personalised, scheduled up to three days ahead and cancellable until it leaves.

twilio-sendgrid-email-settings

Controls how a SendGrid email is put together and what gets counted — reusable layouts filled in per recipient, open and click tracking, and tracked links that show your own domain.

twilio-sendgrid-engagement-quality

SendGrid rates how much recipients actually want your email, one to five, and splits the rating into five parts so you can see which one is dragging it down.

twilio-sendgrid-inbound-parse

Mail sent to your own domain can be handed straight to your software instead of sitting in a mailbox — the basis of turning emails into support tickets automatically.

twilio-sendgrid-suppressions

SendGrid keeps a do-not-send list of addresses that bounced, complained or unsubscribed; this is how to read it, take someone off it, and let people opt out of one kind of mail only.

twilio-sendgrid-webhooks

Sending an email only means SendGrid accepted it; this is how to be told afterwards what actually became of it — delivered, bounced, opened, or reported as spam.

twilio-sms-send-message

Texts that get filtered away, error codes nobody explains, picture messages that only work in three countries — the SMS detail you need once plain sending is not enough.

twilio-taskrouter-routing

A call or a ticket comes in and somebody has to take it — this decides who, by what each person can handle and how busy they already are, instead of you writing that logic yourself.

twilio-verify-send-otp

Sends the short code people type in to prove a phone number or an email address is theirs, and checks the one they type back — by text, a call, email or WhatsApp.

twilio-voice-conversation-relay

Puts an AI on the phone: Twilio turns the caller's speech into text and the AI's reply back into a spoken voice, so your own code only ever deals in text.

twilio-voice-outbound-calls

A phone call placed by software instead of by a person — with control over what the other end hears, who else gets joined to the line, and whether it all gets recorded.

twilio-voice-twiml

While a call is in progress Twilio keeps asking your server what to do next, and the answer is written in a small instruction language — say this, wait for a keypress, transfer, record.

twilio-webhook-architecture

Every Twilio product calls back into your server when something happens, and anyone on the internet can imitate those calls — this is how to prove one is genuine, reply in time, and survive the volume.

twilio-whatsapp-manage-senders

Getting a phone number formally registered as a business account on WhatsApp through Twilio, and keeping it in working order afterwards — including doing it on a client's behalf.

twilio-whatsapp-send-message

WhatsApp only lets a business write freely for 24 hours after the customer last wrote; after that it has to be a pre-approved message, and sending the wrong kind fails silently.

typescript-data-visualization-engineering

Charts that have to live inside a real application: which drawing library suits the job, what shape the data should arrive in, and how to keep the code from tangling as it grows.

ucp

Shopping done from the command line — search thousands of merchants for an item, put it in a basket, pay for it, then check where the order has got to.

ui-toolkit

Drops a ready-made video-call screen into your own web app — the join button, the chat, the captions and the recording already built — running on Zoom's video technology underneath.

ukb-topmed-phewas-skill

Take one spot in the human genome and get back the traits and diseases it has been linked to in a large UK health study, however you happen to name that spot.

uml-and-software-architecture-visualization

Drawing how a system is put together and how it behaves — picking the kind of diagram that answers the question being asked, for the people who will actually have to read it.

uniprot-skill

Looks things up in UniProt, the public reference catalogue of proteins, and gives back a short readable answer instead of dumping the whole record into the conversation.

upgrade-stripe

Stripe changes its payment interface on a dated schedule and old code eventually stops working — this is how to move to a newer version and find out what will break first.

upgrading-expo

Walks an app built with Expo up to a newer version of the framework, and sorts out the breakage that comes with it.

use-dom

Web code — a chart, a rich text editor, anything built for the browser — runs unchanged inside a phone app, hosted in a small embedded browser.

using-git-worktrees

Before new work starts, the project gets its own separate copy to work in, so nothing that happens next can disturb the work you already have open.

using-superpowers

A house rule for an AI assistant: look for a skill that covers the task and follow it before doing anything else — before answering, and before even asking a question back.

v0-dev

Vercel's v0 turns a written description of a screen into working web-app code; this covers driving it from the command line, from your own program, or straight through GitHub to a live site.

validation

Takes the list of possible security bugs a scan turned up and tries to actually make each one happen, so the real ones are proved and the rest are marked unproven with the missing proof spelled out.

vercel-agent

Built into the Vercel hosting platform rather than installed as a package, this AI reads your code changes, chases down why a live site started misbehaving, and wires in its analytics tools for you.

vercel-api

Gives the assistant a live look into your Vercel account — what has been deployed, what failed and why, which settings and domains are actually in place — instead of guessing from the code.

vercel-cli

The terminal command that puts a site live on Vercel and manages it afterwards — plus the linking step people most often get wrong when one repository holds several projects.

vercel-firewall

Decides which visitors a Vercel-hosted site turns away — bad addresses, abusive request rates, suspected bots — and stages each rule so you can watch it against real traffic before it blocks anyone.

vercel-flags

Switches that turn a feature on for some visitors and off for others, so a change can go to a small share of users first — or two versions can run side by side — without shipping new code each time.

vercel-functions

Vercel runs your back-end code for you; this is how to pick where it runs, how long it is allowed to run, and how to make it stream output, hold a live connection, or fire on a schedule.

vercel-queues

A waiting line for work a site does not need to finish while the visitor waits — each job is held safely until something picks it up, and tried again if that something fails.

vercel-sandbox

Some jobs need a real web browser rather than a plain web request. This drives one on a throwaway machine that Vercel starts on demand and destroys when the job is done.

vercel-services

Puts several separate pieces of an app — say a Python back end and a JavaScript front end — on one Vercel site, each one answering a different part of the web address.

vercel-storage

Where a Vercel app keeps its data — uploaded files, small config read on every request, or a full database from the marketplace — and how to get it connected without copying credentials by hand.

verification

Follows a feature all the way through — the page, the request it sends, the data behind it, and what comes back — and stops at the first place it actually breaks, with the evidence.

verification-before-completion

No 'it should work now': the test, the build or the fix has to be run and its real output read before the work can be called done.

view-refactor

Tidies up the screen code of a Mac app built with SwiftUI: one huge file becomes small pieces named after what they do, with the older AppKit parts kept to the edges.

visualization-strategy-and-critique

Works out what a chart should actually show before anyone draws it — which comparison matters, and which form makes it easiest to read — and gives an existing one a straight critique.

web-3d-asset-pipeline

Gets a 3D model ready to ship in a browser game — cleaned up, exported in the one format the web agrees on, and shrunk so the page is not left waiting on it.

web-game-foundations

The decisions to make before writing a browser game: which engine, what the game rules own versus what the screen owns, how the controls map, and where saved data lives.

web-perf

Measures how slowly a page really loads in a real browser, then names the specific things holding it up and how much time each fix would win back.

website-to-hyperframes

Hand it a web address and it makes a short video about that site — an ad, a product tour, a teaser — narrated, and using the site's own colours, fonts and images.

window-management

Makes each window of a Mac app behave the way that particular window should — where it opens, whether it can be resized, whether it comes back next launch, how much title bar shows.

workers-best-practices

Writes or reviews code for Cloudflare Workers against the rules Cloudflare publishes today — the guidance is fetched fresh each time rather than recalled from memory.

workflow

Long jobs that must not be lost halfway: every step is recorded as it completes, so the work picks up where it stopped after a crash, a retry, or a wait for someone to approve it.

wrangler

Wrangler is how code written for Cloudflare gets from your laptop onto the internet: putting it live, attaching storage to it, handling its secrets, and undoing a bad release.

writing-plans

Turns a rough spec into a written plan of small tasks — each one naming the files to touch, the test to write first, and the point where you stop and commit.

writing-skills

Writing a skill an AI assistant will actually follow — tested the way you would test code, to see whether it changes what the assistant does or just gets skipped.

zoom-apps-sdk

Building an app that lives inside Zoom itself — a panel beside a live meeting, an overlay on the video, a tool in the main window.

zoom-cobrowse-sdk

Co-browsing built into a company's own website: during a support call the agent sees the very page the customer is on, live in the browser.

zoom-general

Zoom has a dozen overlapping developer products; this works out which one actually owns the thing you want to build, and how signing in works for it.

zoom-meeting-sdk-android

Zoom meetings running inside an Android app — the sign-in signature, the join-or-start call, and the meeting screen itself, whether that is Zoom's own or one you build.

zoom-meeting-sdk-electron

Electron apps reach the same native Zoom meeting code that desktop apps use, through a Node add-on — so what works depends on the platform underneath, not on the browser.

zoom-meeting-sdk-ios

An iPhone or iPad app that can drop someone straight into a Zoom meeting, or start one as the host, without handing them off to the Zoom app.

zoom-meeting-sdk-macos

A Mac app that opens a Zoom meeting in its own window — joining or hosting from inside the app, with no separate Zoom window in the way.

zoom-meeting-sdk-react-native

One React Native codebase sitting over two native Zoom SDKs — the official wrapper package, and the seams where the JavaScript side meets the native one.

zoom-meeting-sdk-unreal

A Zoom meeting running inside an Unreal Engine project, driven from C++ or from Blueprints, on top of Zoom's native meeting code.

zoom-meeting-sdk-web

Zoom meetings on a web page come in one of two shapes, and this is mostly about picking the right one — then getting the browser's request to join signed by your own server.

zoom-meeting-sdk-web-client-view

The quick way to put Zoom on a web page: the meeting takes over the whole page in the same interface people already know from joining Zoom in a browser.

zoom-meeting-sdk-web-component-view

A real Zoom meeting sitting in a box on your own page, positioned and styled to fit the rest of the app rather than taking over the screen.

zoom-meeting-sdk-windows

Windows software written in C++ that joins Zoom meetings by itself — either showing a window of its own design, or none at all, as a bot that quietly listens and records.

zoom-oauth

Signing in to Zoom from your own software: which of Zoom's several sign-in routes yours should take, how to keep the tokens alive afterwards, and what its refusals actually mean.

zoom-rtms

A live feed of what is happening in a Zoom meeting or a Contact Center call — the audio, the video, the chat, the running transcript — arriving at your own server as it happens.

zoom-video-sdk-flutter

Video calling of your own design inside a Flutter app, running on Zoom's video engine — these are your own sessions, not Zoom meetings.

zoom-video-sdk-linux

On Linux the video never arrives as a ready-made window — a program gets the raw picture and sound and decides what to do with them, whether that is a recording bot in a container or a desktop app that draws its own screen.

zoom-video-sdk-unity

Live video between people, running inside a Unity game or app on Zoom's video engine, with the session's comings and goings wired to what happens in the scene.

Zotero

Reaches into the Zotero reference library on your own computer — finding the paper, pulling the bibliography out, and dropping the citation into the draft you are writing.