/
64 guides
- Claude Code dictation: talk instead of typing, and the transcription doesn't eat your limitsClaude Code has built-in dictation: hold the space bar, talk, and the words show up in the prompt. How to turn it on, how to set your language, how to throw away a wrong sentence, and three sentences to dictate right away.
- pixelfx: 244 ready-made pixel art effects, and hits that finally land244 free pixel art effects made in code, plus a Claude Code skill that uses them and makes new ones. How to install it, the requests to copy, how to put them in your game, how to recolor them and how to make hits land.
- archify: a clickable map of the code Claude wrote for youarchify is a free add-on for Claude Code: with one sentence Claude reads your project and archify draws its map, an HTML file you explore with a click and send to anyone. The install, three requests to copy and the keys to explore the map.
- ponytail: the same work with half the codeponytail is a free add-on for Claude Code: before writing, it looks for what already exists, then writes only the code that's needed. The install, the three intensity levels, the two review commands and the prompts to copy.
- grill-me: it asks you the questions firstgrill-me is a free skill that makes Claude Code ask you questions before it works, one round at a time, each with the answer it recommends. The two ways to install it, a real round of questions of mine and the prompts to copy.
- Advisor: Opus only when it countsClaude Code's advisor pairs a lighter model with a stronger consultant, such as Opus, for the moments that matter. The three ways to turn it on, the six model pairings, what it costs and prompts to copy.
- Impeccable: the site that doesn't look AI-madeImpeccable is a free, open-source add-on that gives Claude Code a design vocabulary and a detector that finds the tells of an AI-made site. How to install it, how to check your site in a minute with no AI, and the prompts to copy to fix the page you already have.
- Do AI assistants recommend you or your rival?With Limelit Open, a free open-source project, you ask ChatGPT, Claude, Perplexity and Gemini your customers' questions and see who they name, in what position and which sites they cite. Then you ask Claude Code about it in plain words.
- When Claude uses the old versionClaude sometimes writes code for an old version of your libraries. Here I explain why it happens and how to use stalemap, an open-source community project, to tell Claude what changed before it starts and to check its work after every edit.
- Where they lookBefore you launch, find out where eyes land on your page: a free tool that runs in your browser, how to read the first-second heatmap and how to fix the page with Claude Code.
- Claude Code from your phoneYour computer is at home, you're not: with Remote Control you start and steer a Claude Code session from the Claude app on your phone. Requirements, the three ways to turn it on, photos, notifications and limits.
- Your project becomes a Mac appHow to turn the web tool you built with Claude into a Mac app with its own icon and window, using the open-source project app-it: what you need, what to type, what it adds and how to remove it.
- You write the lines that matterClaude Code's Learning style, step by step: how to turn it on, what it does when it reaches a real decision, the prompts to use it on a feature of your app and how to switch back when you're in a hurry.
- Good taste for Claude CodeTaste Skill is a free, open-source add-on that makes Claude Code design your app's page the way a designer would. How to install it, how it works, and prompts to copy for a new page, a page to redo and a dashboard.
- One click and Claude Code is already in the projectHow to build a claude-cli:// link that opens Claude Code in the right folder with the request already typed, and where to put it: in an error alert, on your numbers page, in a document or in an alias.
- Ears for Claude Codeearshot is a free GitHub project that lets Claude Code listen to your videos: what you hear second by second, which hits land off time, what to fix first. How to install it, the prompts to copy and the limits to know.
- A page instead of the terminalClaude Code artifacts: real pages on claude.ai, built from the work of your session. What you need, what to ask, how to find them again, share them and keep them updated with a single link.
- The rejection before Apple'sGreenlight reads your app and tells you what could get it rejected by Apple or Google Play, before you submit it. How to install it, how to read the results, how to have Claude Code use it and what it can't see.
- The arrow on the screenClaude Code pointing at the right button with a huge arrow, instead of writing “click Allow”: installation, the Accessibility permission, the hook and the prompts to copy.
- Claude picks up on its ownHow to get Claude Code to pick up on its own when your usage limit resets: the checkbox in the desktop app, the automatic wait in the terminal, when it doesn't work and how to set up a long job for the night.
- Claude is your press officeNewsjack step by step: how to install it in Claude Code, how to tell it once who you are, how it reads the last 24 hours of news every morning and how you go from a story to a checked pitch, one journalist at a time. With its limits, stated plainly.
- A skill makes your App Store screenshotsThe free app-store-screenshots skill for Claude Code, step by step: how to install it, what to ask it, how to use the editor it builds for you, how to take clean captures from the simulator and export the sizes for the App Store and Google Play.
- A thousand fake customers before launchMiroFish step by step: what it is, how to install it on your computer, what to give it to read to rehearse a launch and how to question the fake customers. With its limits, stated plainly.
- A team of hackers against your appHow to run the official Claude Security plugin on your app before launch: what you need, how to install it, how to read the report and apply the fixes. Plus the four typical holes in AI-built apps.
- Message your Claude from TelegramYou connect the Claude Code session open on your computer to Telegram, Discord or iMessage: you text it from your phone, it works on the project's real files and answers you in the chat. And you can send error alerts into that same session.
- Ask your data where people give upHow to connect PostHog to Claude Code through its MCP server, find the step where you lose the most users, change that screen and test it on half your users. With prompts to copy and the rules for not fooling yourself.
- The dashboard with no backendA page with your app's numbers, published by Claude Code with a link you can share: how to connect the data, what updates on its own and what doesn't, who can see it and how Claude answers comments.
- An iPhone inside ClaudeThe iOS simulator inside the Claude desktop app: Claude builds your app, opens it next to the chat and tests it on its own. Plus the demo-screen trick for taking real screenshots from the latest commit, for posts and for the App Store.
- One post, six social networks, one sentenceHow to publish and schedule the same post on TikTok, Instagram, YouTube, X, Threads and LinkedIn by typing one sentence to Claude Code: which scheduler to pick, how to connect it and the real limits of each platform.
- The stranger testBefore I publish a text, I give it to an agent that knows nothing, on the smallest model: three questions and the list of what it didn't understand. The prompt, the agent and the skill to copy, and how to read the answers.
- The site with no APIHow I let Claude Code drive my Chrome on dashboards that have no API: what to install, site permissions, and the procedure written once that the agent follows every time.
- Claude edited this videoVideos made of code with Remotion and Claude Code: the official skills, the flow from script to final file, the frame check, and how to wrap it all in a skill that starts with one command.
- The marketing team that works at nightThe three Claude Code scheduled tasks that analyze, research and produce our app's posts every night: the routine to copy, the critic with its rubric, the stranger test and the rule that stops a run from halting halfway.
- It passes the tests, it isn't safeThe prompt I give a second Claude Code session when the tests are green: I ask it to attack the code that was just written, not to approve it, and I take the report back to the first one.
- How do you know your app is down?The tools I use to know my app is down before my users do, the prompts to connect them in half an hour, and the one to use when the alert goes off.
- The file Claude reads before anything elseMy CLAUDE.md to copy, line by line: what to write in it, where to put it and how to check that Claude Code reads it every session.
- When the chat gets too longThe summary prompt, the decisions file and the commands to start again clean: that way a long chat doesn't make you lose what you'd decided with Claude.
- The workflow: an assembly lineThe five-step chain to paste into Claude Code: search, plan, build, break, summary. How to launch it, follow it and save it as a command.
- The “you handle it” buttonThe rules I use to let Claude Code work freely in the project while blocking deletes, publishing and the real database: the mode, the file to copy and how to check it.
- Go back an hourThe three sentences I tell Claude Code to save the project, go back when something breaks, and try big changes without risk. Plus checkpoints, and what they don't cover.
- The guardTwo hooks for Claude Code: one asks you to confirm before a command deletes files and blocks the irreparable ones, the other runs the tests after every edit.
- The key in the codeThe prompt that searches for keys and passwords in your code and git history, how to move them to .env, and what to do if one has already slipped out.
- Switch off the feature without republishingHow I put a new feature behind a switch I turn on and off from my phone, with no new version: where I keep it, the prompts and the three rules.
- Five thousand apps leaking dataThe two-minute test to find out whether your app is one of the ones leaking data, and the prompt that closes it and tests it again. Only on your own app.
- Why I left LovableHow to move a project out of Lovable or Replit without losing anything: code, database, keys, users, files and domain, with the prompts to use in Claude Code.
- The loopThe three loops I use in Claude Code: the site after a deploy, the tests on the main branch and the reply you're waiting for. With the texts to copy and how to stop them.
- Think first, then touchHow I turn on Claude Code's plan mode, the sentence I use to get a short plan, and how I read it before I say go.
- What happened to LovableHow to get a project out of a platform without losing anything: code, database, users, files and keys, with the prompts to use in Claude Code.
- Show it the screenThe sentence I use so Claude Code tests every change in the browser before it tells me “done,” and the ways to show it the screen.
- One Claude that checks the otherThe prompt I give a second Claude Code session so it hunts for the weak spots in the code that was just written, and how I take its report back to the first one.
- The little helperThe phrases I use to launch Claude Code subagents, how to create a permanent helper that reviews your code, and the cases where it's better not to launch them.
- Does your database answer everyone?The two-minute test to find out whether your app's database also answers people who haven't logged in, and the prompt that locks it. Only on your own app.
- It does it even while you sleepThe two routines I use: a summary of the night every morning at 7:30 and a check of the site every 5 minutes. With the texts to copy and where to set them up.
- Three hundred euros in one nightThe prompt to have Claude put a request limit in your app, and how to set a spending cap and alert in your AI service. Plus the test to see if it works.
- Spend fewer tokensThe five habits I use in Claude Code to spend fewer tokens, with prompts to copy and an example of how the numbers change over a month.
- Jev, the AI that just decidesHow to give Claude Code a check before a destructive command runs: a list of banned commands, a guard hook, and the prompt to have Claude set it up for you.
- The sentence that breaks everythingThe three-line template I fill in before every request to Claude Code, two ready-to-copy examples and the lines I use to check the result.
- Why scaling breaks vibe-coded appsThe prompt to run a test with a thousand fake users on a copy of your project, and the three fixes that almost always solve it: queries, indexes and cache.
- When it says “done” and it isn't trueThe line to put in CLAUDE.md so Claude stops saying “done” without proof, and the four checks I use to make sure it follows it.
- Three jobs at once, zero messHow I get three Claude Code sessions working on the same app at the same time without stepping on each other's toes: one worktree each, with the prompts to copy.
- Don't use ultracode for everythingThe table for choosing how hard Claude Code should think for each job, what ultracode really does, and the three-step rule for not wasting tokens.
- Making Claude work for daysThe logbook file, the sentence that opens every session and an alarm that restarts the work: that's how Claude keeps going on the same project for days without losing the thread.
- The Mac that doesn't sleepThe message I leave Claude before going to sleep, and how I set up the Mac and the permissions so it works all night without doing damage.
- Two Claudes talking to each otherHow I get two Claude Code sessions to talk to each other: how to turn it on, the three ways I use it every day, and the rule that keeps them from stepping on each other's toes.