# Agentmatik - full site content for AI assistants > Agentmatik is an AI automation studio (Prague / remote). We build automated workflows and AI agents that take repetitive work off companies' plates - installed on the client's own accounts, with a free automation audit as the entry point. Key facts: automated workflows (n8n), AI-enhanced workflows, and autonomous AI agents; personal AI assistant installs (OpenClaw, Claude Code, Manus); the client owns every account and credential; languages: English, Czech, German. Contact: matt@agentmatik.ai · Free automation audit: https://agentmatik.ai Pricing: https://agentmatik.ai/pricing.md · Site index: https://agentmatik.ai/llms.txt ## Services - Automated workflows (Level 1): n8n-based automations that run unattended - save 10+ hrs/week. - AI-enhanced workflows (Level 2): automations with AI judgment steps (triage, drafting, extraction) - save 20+ hrs/week. - Autonomous AI agents (Level 3): agents that handle whole processes end to end - save 40+ hrs/week. - Personal AI assistant installs: OpenClaw / Claude Code / Manus set up on your own hardware and accounts. - Apps: Meet Auto Join (live), TubeMD and Skool Video Downloader (launching) - free to use, Pro license $9.99/month unlocks pro features in every app plus partner deal codes. ## Guides (full text) ### The Birthday Champagne AI Video That viral champagne-and-fire clip - with your own face, your outfit, your city. The exact tools, prompts, and fixes, nothing left out. Source: https://agentmatik.ai/guides/birthday-champagne-ai-video ![The final video](/guides/champagne/final-video-v2.mp4) ## What's inside - The one ordering trick that saves you real money on credits - The photo workflow: background, you, small fixes - with every prompt - The 10-second video prompt that makes it look real, not AI - The 6 details that make people stop scrolling (and what breaks the illusion) Made with Nano Banana Pro (the photo) and Seedance 2.0 (the video). About 30-60 minutes. No editing skills needed. ## The one thing that saves you money Most people jump straight to the video. That is where the money goes - making videos burns through credits fast. The people whose clips actually look real do the opposite: get one great photo first, then turn it into the video. Get the photo right and the video almost makes itself. So: **the photo is the real work. The video is the easy part.** ## The whole thing in 5 steps - **1.** Get the tools (Higgsfield is the easy one). - **2.** Add 3 free Claude skills so your prompts come out great. - **3.** Make the photo in Nano Banana Pro - background, then you, then small fixes. - **4.** Turn it into a video in Seedance 2.0 with one prompt. - **5.** Post it. About an afternoon. A few dollars of credits - if you get the photo right first. ## What you need **Higgsfield** if you are starting out - easy to use, and every model you need is in one place. [Kie.ai](https://kie.ai) if you want to run it from Claude Code - more powerful, but you have less control over spending, and video eats credits fast. Two AI models do the work: Nano Banana Pro makes the photo, and Seedance 2.0 turns it into the video. And 3 free Claude skills make your prompts much better - they teach the AI how to prompt these exact tools: [banana](https://github.com/AgriciDaniel/banana-claude) (photo prompts), [seedance-prompt-en](https://github.com/dexhunter/seedance2-skill) (video prompts), [higgsfield-generate](https://github.com/higgsfield-ai/skills) (images and videos in one step). ## Step 1 - The photo (this is 80% of it) One photo of you sitting down - cake on one side, champagne on the other. This becomes the first frame of your video, so it has to look like you and be set up right. ### A) Make the background Pick a spot people will recognize - I used a hill over Prague with the TV tower in the view. Make it in Nano Banana Pro at 9:16, with your real photo of the spot attached. ![Real photo of the Prague lookout](/guides/champagne/real-spot-v2.jpg) ![AI-generated night bench scene](/guides/champagne/generated-bench-v2.jpg) ```prompt Background prompt A quiet empty wooden park bench viewed straight on, centered in the frame, sitting on a brick-paver path with a few scattered autumn leaves in the foreground. Behind the bench the ground drops away to a panoramic night view of Prague: glowing rooftops and warm window lights, and on the right the illuminated Zizkov television tower lit in red, white and blue with a single vertical blue light beam rising straight up into a deep blue dusk sky. Bare tree branches frame the top corners and dark foliage frames both sides. Empty, still, no people. Composition: symmetrical, eye-level, the whole bench visible from armrest to armrest with generous foreground path and headroom above. Shot on a full-frame camera at blue hour, soft ambient city light, gentle warm bokeh from distant streetlamps on the left, natural low-light look. Vertical 9:16. No text, no watermark. ``` ### B) Put yourself in the photo This is the step that makes or breaks it. Attach the background plus 3-4 clear photos of your face (front, side, angle) and your outfit. The thing most people get wrong: keep your face big in the photo (waist-up, not full-body), and use a few clear face photos - not one tiny collage. If your face is small, the AI invents a stranger. Tell it plainly: "a real, specific person - do not make him younger, slimmer or better-looking." ![Empty AI bench scene](/guides/champagne/generated-bench-v2.jpg) ![Matt placed into the scene](/guides/champagne/matt-in-frame-v2.jpg) ```prompt The prompt people usually share online Create a photorealistic birthday luxury lifestyle / night editorial photo. In the frame, a beautiful girl sits on a wooden bench outside at night. She is wearing an elegant black tight midi dress with thin straps and black strappy high heels. Her legs are beautifully crossed, the pose is confident, feminine, and slightly cheeky. The girl has voluminous hair, soft glamorous makeup, glowing skin, nude lips, and a calm, confident look into the camera. In one hand, she holds a bottle of champagne from which foam and splashes are spectacularly flying out. In the other hand, there is a small birthday cake on a stand with burning candles. Location: a night city street or park, dark background, soft lights in the distance. Light: flash like in expensive film photography, bright accent on the face, body, cake, and champagne splashes, the background is darkened. Style: photorealistic, high-end editorial, ultra realistic, night flash photo, detailed skin texture, cinematic contrast. Important: realistic anatomy, natural hands and fingers, no facial distortions, no extra objects, no text, no watermarks, no cartoonish look. ``` ```prompt My photo prompt (the one I used) Photorealistic night photograph, vertical 9:16, MEDIUM shot framed from the knees up (the man's face is large in frame for maximum likeness). A man sits centered on the wooden park bench at night, with the city skyline, the illuminated red-white-blue tower and its vertical light beam behind him (match the attached bench scene). IDENTITY - render his face and head to EXACTLY match the FACE reference photos: same face shape, eyes, nose, thick black rounded glasses (always worn), full dark beard with light-grey flecks, short dark hair greying at the temples, fair skin, solid build. He is a real specific person - do NOT idealize, slim, or de-age him; he must be clearly recognizable. Grayscale tattoo sleeve on the LEFT forearm. WARDROBE - he wears a white tuxedo jacket over a crisp white dress shirt with a black bow tie (take the outfit from the tuxedo reference photo). Tailored, elegant. He has a calm, cool expression - a neutral POKER FACE or a very slight closed-mouth smile - looking straight at the camera. In his LEFT hand (left of frame) a small chocolate cake topped with a SINGLE gold metallic numeral-"5" candle, one lit flame, no other candles. In his RIGHT hand (right of frame) an upright dark champagne bottle. Realistic skin texture, natural anatomy and correct hands, cinematic night flash photography, high detail. Resolution 2K-4K. No text, no watermark. ``` ### C) Small fixes Two quick edits help a lot: lift the cake up so the fire has room to spread over it, and (optional) empty the bench if you want to walk into the shot. I skipped the walk-in - it looked too fake, and I would rather use the 10 seconds on the big moment. ![Before: cake held low](/guides/champagne/matt-in-frame-v2.jpg) ![After: cake lifted](/guides/champagne/cake-lifted-v2.jpg) ```prompt Lift the cake Edit this photo: the man RAISES the chocolate cake (with the gold "5" candle) higher and extends his LEFT arm outward to his left, holding the cake up around chest/shoulder height, away from his body and clearly out to the LEFT side of the frame. Keep the champagne bottle held out to his RIGHT. Keep everything else identical - same face, glasses, beard, white tuxedo, bench, skyline, tower beam, lighting and framing. Photorealistic, vertical 9:16, no text, no watermark. ``` [[newsletter]] ## Step 2 - The video (the easy part) Drop your photo in as the first frame and describe what happens, step by step. Settings: 10 seconds · 9:16 · 1080p · high quality. ```prompt My video prompt (the one I used) @image_1 as the first frame. Keep his exact face, glasses, beard and white tuxedo identical every frame. NO last frame - the clip ends on the action. Vertical 9:16, one continuous take, photorealistic night. Prague hilltop bench; behind him the skyline and the illuminated red-white-blue tower with a vertical light beam. CAMERA: stays at a MEDIUM waist-up framing the whole clip - only the faintest slow drift in. NEVER a tight close-up (that breaks the likeness). Keep his full upper body, the cake and the bottle in frame the whole time. 0-1.5s (real time): He sits with a cool, neutral deadpan POKER FACE - chocolate cake with the gold "5" candle raised up and out to his LEFT at chest height, upright champagne bottle held LOW in his RIGHT hand with its base on the bench seat. The "5" flame flickers. 1.5-2.5s (real time): He firmly PUSHES the bottle base straight DOWN onto the bench seat. The CORK visibly bursts off the top and shoots HIGH up into the frame on the RIGHT - spinning against the dark sky with a puff of vapour, clearly seen. Give the cork its own clear moment. 2.5-3s: Speed ramps down into dramatic SLOW MOTION as champagne begins to surge from the bottle. 3-7.5s (extreme slow motion, almost a freeze at the peak): RIGHT - a tall VERTICAL champagne fountain shoots straight up, foam and droplets glinting in the air. LEFT - a big orange-yellow fireball ignites AT the "5" candle and blooms ACROSS the cake, with rolling flame, dark smoke and rising embers. His head tilts DOWN toward the cake, eyes closed mid-blow, face lit between the two effects. The two effects stay on opposite sides and NEVER cross the centre. 7.5-8.5s (eases back to real time): Fire fades to drifting smoke on the LEFT; champagne weakens to fizzing drips on the RIGHT. 8.5-10s (ending): He holds the deadpan, then breaks into a faint, confident SMIRK - cool and unbothered - as smoke drifts across the tower beam. Continuity: cake with the "5" stays raised in his LEFT hand on the LEFT; bottle base stays on the bench on the RIGHT - never swap or twist. Deadpan throughout (only the blow, then the smirk). Fire starts at the cake and goes away from his face; fire and champagne never mix or cross the centre. Realistic fire and liquid physics only - no sparkles, fireworks, confetti, text, captions or watermark. Audio: night city ambience and wind; a wooden bottle-tap, then a loud cork POP and explosive champagne fizz on the right; a fire WHOOMP and crackle on the left. No speech, no music. ``` ## What makes it look great - **Keep a straight face.** Staying calm while everything explodes around you is the joke. - **Keep the two effects apart.** Champagne on one side, fire on the other - a messy middle looks fake. - **Show the cork fly.** Most clips hide it. Let it pop and shoot up - that is the "wait, how?" moment. - **Slow it right down** at the big moment, then speed back up. - **Don't zoom in too close.** A tight close-up stops looking like you. - **Your first photo is your thumbnail.** Make it genuinely good. ## Two quick pro tips **1) Tag your photos.** In Higgsfield you can tag an image in the prompt so it is clear which one you mean. Right-click an image to set it as the start frame, end frame, or just a reference. **2) Connect Higgsfield to Claude** so you can prompt it straight from chat: Claude → Settings → Connectors → Add Custom Connector → name it Higgsfield, MCP URL https://mcp.higgsfield.ai/mcp → Connect and sign in. No API key needed. ## Before you start - Higgsfield or Kie.ai account ready - 3 skills added: banana, seedance-prompt-en, higgsfield-generate - Background made - Photo of you - looks unmistakably like you - Cake lifted / small fixes done - 10-second video made - Posted - and tagged me Questions, or made one? Tag [@mattzimak.ai](https://www.instagram.com/mattzimak.ai) - I share builds like this whenever there is one worth sharing. --- ### How to structure your Claude Code workspace Setting up Claude Code? A research-backed blueprint: one workspace repo, a 30-minute setup with /init and /permissions, and clear rules for when work earns its own folder or repo. Source: https://agentmatik.ai/guides/claude-code-workspace-structure ## What's inside - The from-scratch blueprint: one workspace repo, an annotated folder tree, and a lean root CLAUDE.md - A 30-minute setup walkthrough with the real commands: /init, /memory, /permissions, /mcp - Growth rules: when work earns a folder - and the 3-question test for when it earns its own repo - How Claude Code really loads context (verified against the official docs), 9 anti-patterns, and the myths to ignore One giant workspace or a repo per project? We pulled 295 sources - the official docs, engineering write-ups, and hundreds of practitioner threads - into one answer. Here is the blueprint we use for our own automation work. You can build it from scratch this afternoon, even if today you have zero repos. ## The short version Create one private workspace repo for everything you work on with Claude Code. Give each domain of work its own subfolder with its own CLAUDE.md as it appears, launch Claude Code from the subfolder for focused sessions and from the root for cross-cutting work, and give something its own repo only when it has a genuinely different lifecycle, audience, or access boundary - a deployable product, a knowledge base with its own review flow, code another team owns, or client work under NDA that needs hard isolation. Per-role or per-function repos almost always fail the same way: you live in one of them, the rest go stale, and the "back everything up to GitHub" goal silently breaks because nobody pushes repos they never open. ## Start here: one private workspace repo On day one you need exactly one repo. Create it, make it private, and treat it as the home for every file you and Claude touch that does not obviously belong somewhere else: ``` mkdir ~/code/workspace cd ~/code/workspace git init ``` Here is what it grows into. You do not create all of this today - the point of the blueprint is knowing where things will go when they appear: ``` workspace/ # ONE private repo ├── CLAUDE.md # lean root: what this is, the map, global rules ├── .claude/ │ ├── settings.json # permissions (deny .env reads), hooks - committed │ └── skills/ # workspace-wide skills ├── .mcp.json # root-scope MCP servers, ${ENV_VAR} refs only ├── .gitignore # .env, big binaries, caches, .claude/worktrees/ ├── automations/ # domain A - your main domain can be the root itself ├── marketing/ # domain B │ ├── CLAUDE.md # scoped: what lives here, conventions │ ├── .mcp.json # domain MCP servers (load when launched here) │ └── .claude/skills/ # domain skills (apply when working here) ├── clients// # per-client folders, own CLAUDE.md each ├── docs/ guides/ plans/ # workspace-wide knowledge └── tools/ scripts/ # shared deterministic tooling ``` Separate repos live alongside it - only what passes the 3-question test below: ``` ~/code/ ├── workspace/ # the monorepo above ├── product-site/ # each deployable product: its own repo ├── team-shared-repo/ # repos other people push to └── knowledge-base/ # SSOT repos with their own review flow ``` ## The setup walkthrough (about 30 minutes) - **1. Launch Claude Code in the folder.** Run **claude** inside ~/code/workspace. The directory you launch from decides which [settings](https://code.claude.com/docs/en/settings) and [project MCP servers](https://code.claude.com/docs/en/mcp) load, so the workspace root is home base from the first minute. - **2. Generate a starter CLAUDE.md with /init.** The [/init command](https://code.claude.com/docs/en/commands) scans the folder and bootstraps a CLAUDE.md for you; set **CLAUDE_CODE_NEW_INIT=1** to get the newer interactive flow that asks about your setup instead of guessing. Treat the output as a draft, not a final file. - **3. Cut the draft down to a map, not a manual.** Aim for well under 200 lines - the [official guidance](https://code.claude.com/docs/en/claude-directory) and the single most repeated community lesson (the best root files are [shockingly small](https://www.humanlayer.dev/blog/writing-a-good-claude-md)). Keep four things: one paragraph on what the workspace is, a table of top-level folders, your global rules (secrets, style, logging), and pointers to deeper docs instead of their content. Refine it anytime with [/memory or the # shortcut](https://code.claude.com/docs/en/memory). A good example: one practitioner runs his whole [personal OS](https://amankhan1.substack.com/p/how-carl-set-up-his-personal-os-in) on a root file under 100 lines. - **4. Lock down secrets with /permissions.** Add [deny rules](https://code.claude.com/docs/en/permissions) on file reads for .env and anything secret-shaped - deny beats allow, and a deny rule is the real protection; an ignore-style exclusion alone is not. Keep the actual .env gitignored, with a secret manager as the source of truth. - **5. Wire tools with /mcp.** Add MCP servers to a project [.mcp.json](https://code.claude.com/docs/en/mcp) so they are versioned with the workspace - and reference credentials as ${ENV_VAR}, never pasted values. - **6. Write the .gitignore before the first big commit.** .env, caches, .claude/worktrees/, and every folder that will ever hold video, design exports, or generated assets - big binaries go to cloud storage, not git. One careless git add -A can try to swallow gigabytes. - **7. Commit and push.** Create the private GitHub remote and push. From now on the workspace is your working memory: treat an unpushed workspace as an incident. [[newsletter]] ## Growing it: folders first, repos only at a boundary ### A second kind of work earns a folder When a genuinely different domain shows up - say marketing next to your build work - it gets a subfolder with its own CLAUDE.md (what lives here, domain conventions, which root rules apply with extra force - additive to the root, never contradictory). If the domain needs its own tools, give it a domain .mcp.json and [directory-scoped skills](https://code.claude.com/docs/en/skills). Then adopt the habit that makes the whole structure pay off: **cd workspace && claude** gives you a root session: the full map, cross-domain work, no domain MCP servers. **cd workspace/marketing && claude** gives you a marketing session: the root CLAUDE.md plus the marketing CLAUDE.md, MCP servers, and skills. Every other domain stays out of the way - which is exactly the lean-context behavior [Anthropic's context engineering guidance](https://www.anthropic.com/engineering/effective-context-engineering-for-ai-agents) pushes for. For parallel sessions on the same repo, use [git worktrees](https://code.claude.com/docs/en/worktrees) - keep them under .claude/worktrees/ and gitignore that folder. And when you need another repo's code in context, **claude --add-dir ../other-repo** beats copying files around. ![Diagram: domains earn folders inside the workspace; a separate repo only when lifecycle, audience, or access says yes](/guides/claude-code-workspace-structure/folders-vs-repos.jpg) ### Something earns its own repo only at a boundary A separate repo is justified only if at least one answer is yes: - **Lifecycle** - does it deploy, release, or version independently? If yes: products, websites, published packages, and browser extensions get their own repo. - **Audience** - do other people (or other AI agents) need access to it? If yes: team repos, client-shared repos, and public projects get their own repo. - **Access boundary** - would mixing it in violate privacy, an NDA, or compliance? If yes: personal data, client-confidential work, and regulated code stay separate. Everything else - notes, plans, guides, scripts, exported configs, research, client working files only you touch - stays in the one workspace repo. ### Why the one-workspace direction wins - **Cross-cutting context is the whole point.** A coding agent can update an API contract, regenerate types, fix the call sites, and test end to end in one session - the reason [AI-era repo guidance](https://www.augmentcode.com/learn/monorepo-vs-polyrepo-ai-s-new-rules-for-repo-architecture) tilts toward monorepos. Split repos force you to ferry context between sessions by hand. - **Lazy loading solves the pollution fear.** Subfolder CLAUDE.md files do not load at launch - they load only when Claude works with files in that subtree ([official behavior](https://code.claude.com/docs/en/memory)). A big workspace does not mean a big context. - **One backup to keep healthy.** Ten repos means ten push disciplines. In practice, the repos you do not open daily rot unpushed. - **Practitioners who tried both agree.** Teams that made their monorepo agent-ergonomic report [large velocity gains](https://www.getbasis.ai/blogs/how-we-made-our-monorepo-ergonomic-for-agents); the strongest counter-arguments are about huge multi-team codebases ([CI cost, team autonomy](https://code.claude.com/docs/en/large-codebases)) - not about a personal or agency workspace. ## How Claude Code actually loads context This is the "why" behind the blueprint - and several popular blog claims about it are wrong (see the myth-busting section at the end). Every point links to the doc that verifies it. - **1.** At launch, Claude Code loads your global ~/.claude/CLAUDE.md plus every CLAUDE.md from the launch directory up the ancestor chain - always in context ([memory docs](https://code.claude.com/docs/en/memory)). - **2.** Subfolder CLAUDE.md files lazy-load: they enter context only when Claude reads or edits files in that subtree ([memory docs](https://code.claude.com/docs/en/memory)). This is what makes a large workspace cheap. - **3.** Memory files concatenate - they do not override each other ([memory docs](https://code.claude.com/docs/en/memory)). Write nested files so they add scope-specific detail instead of contradicting the root. - **4.** .claude/settings.json is not inherited from parent directories - it loads from the directory you start Claude in ([settings docs](https://code.claude.com/docs/en/settings)), so a subfolder session does not get the root folder's settings automatically. - **5.** .mcp.json is read from the launch directory too ([MCP docs](https://code.claude.com/docs/en/mcp)). A marketing/.mcp.json loads when you start Claude inside marketing/ - not when you start at the workspace root. - **6.** Nested .claude/skills/ folders do apply when Claude works inside that subfolder, even if the session started at the root ([skills docs](https://code.claude.com/docs/en/skills)). - **7.** additionalDirectories in settings grants file access only - it does not load that directory's CLAUDE.md or skills ([settings docs](https://code.claude.com/docs/en/settings)). - **8.** Permissions: deny beats allow, and a deny rule on file reads is the real protection for secrets ([permissions docs](https://code.claude.com/docs/en/permissions)). - **9.** Session transcripts are stored in plaintext under ~/.claude/projects/ ([.claude directory anatomy](https://code.claude.com/docs/en/claude-directory)). Anything Claude reads can end up there - this matters for client isolation (see the tiers below). - **10.** claudeMdExcludes exists to skip specific CLAUDE.md files in very large monorepos ([large-codebases docs](https://code.claude.com/docs/en/large-codebases)). ## Git and backup rules that save you later - **Private repo, push discipline.** The workspace is your working memory - treat an unpushed workspace as an incident. Multi-repo setups fail exactly here. - **Secrets never in git.** Env-var references in [.mcp.json](https://code.claude.com/docs/en/mcp), a gitignored .env as the local cache, a secret manager as the source of truth - and a pre-commit secret scan such as gitleaks. - **Big binaries never in git.** Videos, design assets, and exports go to cloud storage; gitignore those folders. One careless git add -A can try to swallow gigabytes. - **Archive, do not delete.** When retiring old repos, archive them on GitHub - the history stays readable for free and every change is reversible. - **History is a leak surface.** When folding an old repo's content in, import content only (rsync, not a git merge) if its history ever contained secrets - and rotate anything that was committed. These are not hypothetical risks. In one workspace we audited, 44 production skills existed only inside an uncommitted git worktree - one prune command away from disappearing, and invisible to the "everything is on GitHub" backup story. In another, a pre-commit scan caught 16 live tokens sitting in exported workflow configs before they reached GitHub. Exported configs, not source code, are the classic leak vector. ## Client work: three isolation tiers The one place where "just use the workspace" needs qualification. Tier your clients by sensitivity: - **Tier 1 - normal client work (the default).** A clients// folder inside the workspace with its own CLAUDE.md and secrets via env references. Fine for most agency work where you are the only operator. - **Tier 2 - confidential (NDA, competing clients).** A separate private repo per client. Launch Claude only from inside it, with a per-repo [.mcp.json](https://code.claude.com/docs/en/mcp) so client credentials never load anywhere else. And remember the plaintext transcripts: a root-launched session can read every client folder, so do not start root sessions if browsing across clients would already be a breach. - **Tier 3 - regulated or high-sensitivity.** A dedicated devcontainer or VM per client with allowlisted egress; nothing shared. Mixed-tool teams: if collaborators use other coding agents too, keep the canonical instructions in AGENTS.md and make CLAUDE.md a one-line import of it ([memory docs](https://code.claude.com/docs/en/memory)). ## Already have work scattered across several repos? Then consolidate content, not history: rsync the files into workspace subfolders (scan each old repo for committed secrets first - findings mean rotate the keys and never merge that git history), then archive the old repos on GitHub instead of deleting them, so everything stays reversible. And if you have many live code repos you cannot merge - deployed products, CI you must not touch - skip the merge entirely: a [virtual monorepo](https://medium.com/devops-ai/the-virtual-monorepo-pattern-how-i-gave-claude-code-full-system-context-across-35-repos-43b310c97db8) (a parent folder cloning them by domain, with a top-level CLAUDE.md system map) or an identical "repository ecosystem" table in every repo's CLAUDE.md gives Claude the full-system view without moving anything. ## Nine anti-patterns we keep seeing - **The dumping-ground CLAUDE.md.** One 900+ word root file mixing conventions, project state, and reminders. Signal-to-noise collapses and the instructions get ignored ([keep it small](https://www.humanlayer.dev/blog/writing-a-good-claude-md)). Split it and push detail down. - **Per-role repos.** The role you do 90% of the time absorbs everything; the rest die unpushed. Domains earn folders, not repos. - **Over-engineered .claude/ scaffolding.** Elaborate rules, commands, and subagent trees built up front rot fast - one practitioner deleted 60% of theirs within four months. Add structure when a pain repeats, not before ([best practices](https://code.claude.com/docs/en/best-practices)). - **Secrets in .mcp.json or exported configs.** The most common real leak. Env references plus pre-commit scanning. - **Unignored binary folders.** One git add -A away from an 11 GB commit. - **Root-launched everything.** Never using subfolder sessions means every session pays for every domain's context and MCP servers - the opposite of [effective context engineering](https://www.anthropic.com/engineering/effective-context-engineering-for-ai-agents). - **Trusting "deeper CLAUDE.md overrides shallower".** It concatenates ([memory docs](https://code.claude.com/docs/en/memory)). Contradictory instructions produce coin-flip behavior. - **A git repo in your home directory.** It shadows every repo beneath it and confuses tooling. - **Mirroring live systems into the repo as docs.** Exports go stale and leak tokens. Declare the live system the source of truth and export deliberately. ## Myths that failed verification Popular claims we checked against the official docs and could not confirm - do not build your structure on these: - **"A deeper CLAUDE.md takes priority over the root on conflict."** Wrong - the [memory docs](https://code.claude.com/docs/en/memory) describe concatenation, with no override semantics. - **"CLAUDE.local.md loads last, so it wins."** A [deprecated mechanism](https://code.claude.com/docs/en/memory) - use imports of untracked files instead. - **"The .claude folder controls about 80% of Claude Code's capability."** An invented statistic - the [.claude directory docs](https://code.claude.com/docs/en/claude-directory) make no such claim. - **"CLAUDE.md is often ignored, so inject everything via hooks."** Overstated - a lean, well-structured CLAUDE.md is followed ([best practices](https://code.claude.com/docs/en/best-practices)). Hooks are for enforcing hard rules, not a workaround for bloat. - **"Claude Code has no project boundaries, so you need an external memory tool."** Overstated - workspace trust, [launch-directory settings](https://code.claude.com/docs/en/settings), and directory-scoped CLAUDE.md files are exactly those boundaries. --- ## Curated agent skills (Claude Code / OpenClaw) Directory with install commands and star counts: https://agentmatik.ai/skills - **agents-skills-autoreview** (Development): Automatic review of agent skills before they run in an OpenClaw instance - the marketplace's guardrail against installing something you did not read. · Why: If you run OpenClaw with community skills, an auto-review layer is the minimum sensible protection. - **AI Project Starter** (Meta, 8★): Opinionated project scaffold with a pre-configured CLAUDE.md, folder structure, and best-practice defaults to start any Claude Code project cleanly. · Repo: https://github.com/jlaran/AI-Project-Starter-Claude - **ai-marketing-skills** (Marketing & SEO, 2.8K★): Open-source marketing skills - growth experiments, sales pipeline, content ops, outbound, SEO, and finance automation. · Repo: https://github.com/ericosiu/ai-marketing-skills - **antfarm** (Orchestration, 2.5K★): Free, open-source multi-agent orchestration on OpenClaw - YAML-defined pipelines with a fresh context per AI agent, and pre-built workflows for feature development, security audits, and bug fixes. · Why: The cleanest way we have seen to run a team of OpenClaw agents without extra infrastructure. · Repo: https://github.com/snarktank/antfarm - **banana-claude** (Creative AI, 818★): Turns Claude into a Nano Banana Pro creative director - optimized prompts with the 5-component formula, image generation and editing via the Gemini API, inpainting, style transfer, 4K. · Repo: https://github.com/AgriciDaniel/banana-claude - **canvas-design** (Webdesign, 159K★): Anthropic's official skill for creating visual art, posters, and compositions as real PNG or PDF files directly from Claude Code. · Repo: https://github.com/anthropics/skills - **caveman** (Token efficiency, 86K★): Cuts around 65% of token usage by instructing Claude to answer in compressed shorthand - for long sessions where context budget matters. · Repo: https://github.com/JuliusBrussee/caveman - **claude-ads** (Marketing & SEO, 6.8K★): A full Meta Ads skill pack - competitive intelligence (/spy), bulk copy generation, a 186-check account audit with health score, and 6-dimension ad scoring. · Repo: https://github.com/AgriciDaniel/claude-ads - **claude-mem** (Memory, 86K★): Persistent context across sessions - captures what Claude does, compresses it, and injects the relevant parts back into future sessions via hybrid search. · Repo: https://github.com/thedotmack/claude-mem - **claude-obsidian** (Research, 8.9K★): A persistent, compounding wiki vault for Claude based on the LLM Wiki pattern - /wiki, /save, and /autoresearch build a self-updating knowledge base. · Repo: https://github.com/AgriciDaniel/claude-obsidian - **claude-video-vision** (Creative AI, 871★): Lets Claude watch and understand videos - extracts frames via ffmpeg, processes audio, and produces creative briefs, hooks, and shot-by-shot breakdowns. · Repo: https://github.com/jordanrendric/claude-video-vision - **code-reviewer (jeffallan)** (Development): Structured pre-PR review: what is broken, what is risky, what is messy, and what is solid. - **codex-plugin** (Meta, 26K★): Use OpenAI Codex from within Claude Code - review code with a second model, run tasks in parallel, or delegate subtasks to a separate coding AI agent. · Repo: https://github.com/openai/codex-plugin-cc - **competitive-ads-extractor** (Marketing & SEO, 67K★): Analyzes 3-5 competitors' active ads and outputs a gap analysis - hook frequency, offer structures, CTA patterns, and untapped emotional angles. · Repo: https://github.com/ComposioHQ/awesome-claude-skills - **content-research-writer** (Content, 67K★): Full research to outline to draft pipeline - adds citations, improves hooks section by section, and enables systematic content production. · Repo: https://github.com/ComposioHQ/awesome-claude-skills - **debug-skill** (Development, 296★): Real breakpoints inside Claude Code - step through execution line by line and inspect live variable state when stuck on a bug. · Repo: https://github.com/AlmogBaku/debug-skill - **design-critique** (Webdesign, 22K★): Structured UX/UI critique from Anthropic's knowledge-work plugins - evaluates usability, hierarchy, clarity, and interaction quality with concrete improvements. · Repo: https://github.com/anthropics/knowledge-work-plugins - **dont-hack-me** (Security): A security self-check for OpenClaw installs - audits your agent's configuration for the common gaps (exposed credentials, over-broad permissions, missing sandboxing). · Why: Every self-install we have audited had at least one gap this catches for free. - **emil-design-eng** (Webdesign, 5.2K★): Emil Kowalski's design-engineering principles as a skill - guides Claude to improve animations, transitions, spacing, and component behavior so products feel genuinely polished. · Repo: https://github.com/emilkowalski/skill - **feature-forge** (Development): Run before writing any code - asks the right questions, thinks like a PM and a developer, and hands you a full spec with acceptance criteria. - **figma-skills** (Webdesign, 1.7K★): Official Figma skills: read Figma files, extract design tokens, and implement designs with pixel accuracy through the Figma MCP server. · Repo: https://github.com/figma/mcp-server-guide - **find-skills** (Meta, 25K★): Discovers and installs skills from the open ecosystem - checks the skills.sh leaderboard, ranks by installs and source reputation, and installs straight from GitHub. · Why: The package manager habit: instead of building from scratch, we check what already exists and vet it. · Repo: https://github.com/vercel-labs/skills - **find-skills (ClawHub)** (Meta): Auto-discovers and installs OpenClaw skills on demand from ClawHub - ask for a capability and it finds, vets, and installs the skill. · Why: Same habit as on Claude Code: check what exists before building. This is the OpenClaw equivalent. - **framer-motion-skill** (Webdesign, 2★): Production-grade animation patterns for React with Motion - scroll animations, exit animations, stagger variants, accessibility via reduced motion, and ready-made recipes. · Repo: https://github.com/Schoepplake/framer-motion-skill - **frontend-design** (Webdesign, 159K★): Anthropic's official frontend design skill - encodes web interface guidelines, component patterns, and accessibility rules so the AI agent produces production-grade UI by default. · Why: Raises the floor: UI output goes from generic to shippable without extra prompting. · Repo: https://github.com/anthropics/skills - **google-ads-skills** (Marketing & SEO, 30★): Twelve Google Ads skills covering campaign audits, keyword work, and reporting. · Repo: https://github.com/itallstartedwithaidea/agent-skills - **higgsfield-generate** (Creative AI, 523★): One-command generation across 30+ models - Nano Banana for images, Seedance for video with native audio, plus reframe and virality scoring - through the official Higgsfield integration. · Why: Powers our creative AI guides, including the champagne video - one skill instead of five tabs. · Repo: https://github.com/higgsfield-ai/skills - **huashu-design** (Webdesign, 21K★): Turns one sentence into finished prototypes, HTML slide decks, MP4 animations, or infographics - 20 built-in design philosophies, verified in a real browser via Playwright before delivery. · Repo: https://github.com/alchaincyf/huashu-design - **humanizer** (Content, 28K★): Removes the signs of AI-generated writing from any text - em-dashes, hype vocabulary, rule-of-three stacking, generic conclusions - based on Wikipedia's documented list of AI tells. · Why: Every piece of copy on agentmatik.ai passes through it. Non-negotiable in our content pipeline. · Repo: https://github.com/blader/humanizer - **hyperframes** (Creative AI, 33K★): HeyGen's declarative HTML/CSS video frames that render to MP4 - purpose-built for AI agents producing video without a GUI. · Repo: https://github.com/heygen-com/hyperframes - **impeccable** (Webdesign, 44K★): A design auditor enforcing 27 deterministic anti-pattern rules - no cheesy gradients, no cards-inside-cards - plus 23 slash commands for auditing and polishing frontend work. · Why: Catches AI-slop design before it ships; we run it on every UI change. · Repo: https://github.com/pbakaus/impeccable - **landing-page-design** (Webdesign, 588★): Higher-converting landing pages - improves structure, messaging hierarchy, CTAs, and visual layout, balancing aesthetics with conversion. · Repo: https://github.com/inference-sh/skills - **larry (LarryLoop)** (Creative AI): TikTok video-making skill born on LarryBrain, now a standalone app - scripts, cuts, and publishes short-form video. - **last30days** (Research, 50K★): Researches any topic across Reddit, X, YouTube, and Hacker News from the last 30 days, then synthesizes the findings into usable prompts and summaries. · Why: Our content research starts here - it is how the guides stay grounded in what people actually struggle with this month. · Repo: https://github.com/mvanhorn/last30days-skill - **linear-skill** (Project management, 116★): Manage Linear issues, projects, and teams from Claude Code - MCP tools, SDK automation, and GraphQL patterns. · Repo: https://github.com/wrsmith108/linear-claude-skill - **magic-mcp** (Webdesign, 5.3K★): 21st.dev's Magic MCP server - generates polished, production-ready UI components from natural language, with logo integration. · Repo: https://github.com/21st-dev/magic-mcp - **mobile-apps** (Development, 437★): Build and test mobile apps from Claude Code - React Native and Expo workflows. · Repo: https://github.com/sleekdotdesign/agent-skills - **nano-banana-2-skill** (Creative AI, 391★): AI image generation on Gemini 3 Pro - green-screen transparency, reference images, and style transfer inside Claude Code. · Repo: https://github.com/kingbootoshi/nano-banana-2-skill - **obsidian-skills** (Meta, 40K★): Teaches Claude Code to natively create and edit Obsidian formats - Markdown with wikilinks, Canvas files, Bases - and drive the Obsidian CLI. · Repo: https://github.com/kepano/obsidian-skills - **openclaw-supermemory** (Memory, 789★): Unlimited persistent memory for an OpenClaw agent - everything it learns survives restarts and long gaps, retrievable by meaning rather than keywords. · Why: An assistant that forgets last month is a chatbot. Memory is what makes it feel like staff. · Repo: https://github.com/supermemoryai/openclaw-supermemory - **paid-media-skills (Ryze)** (Marketing & SEO): Seven OpenClaw skills for paid media - Google and Meta ads workflows from the get-ryze guide. - **playwright-skill (TDD)** (Development, 306★): Walks the full test-driven cycle before implementation: write the failing test, watch it fail, write just enough to pass. · Repo: https://github.com/testdino-hq/playwright-skill - **prompt-guard** (Security): Advanced prompt-injection defense for OpenClaw - screens incoming content (emails, web pages) for instructions that try to hijack your agent. · Why: An agent that reads your inbox will eventually read a malicious email. This is the seatbelt. - **qmd-skill** (Token efficiency, 692★): Cuts an OpenClaw agent's token usage by up to 95% by compressing context handling - the difference between a cheap always-on assistant and an expensive one. · Why: Always-on agents burn tokens around the clock; this is the single biggest cost lever we know. · Repo: https://github.com/levineam/qmd-skill - **rag-architect** (Development): Designs the full RAG pipeline for vector search and knowledge-base projects: chunking, embeddings, retrieval, reranking. - **remotion-skills** (Creative AI, 3.9K★): Programmatic video with React and Remotion - animations, timing, and audio sync knowledge for production-grade output. · Repo: https://github.com/remotion-dev/skills - **secure-code-guardian** (Security): Writes secure code from the start for anything touching auth, passwords, or user input. - **seedance-prompt-en** (Creative AI, 2.5K★): Turns a rough video idea into a precise, production-ready Seedance 2.0 prompt - reference syntax, camera language, beat-matching - built from ByteDance's official docs. · Why: Pure prompt craft, no API key needed - the difference between a generic clip and a usable shot. · Repo: https://github.com/dexhunter/seedance2-skill - **Skill Seekers** (Meta, 14K★): Converts any documentation site, GitHub repo, or PDF into a ready-to-use Claude skill, with automatic conflict detection. · Repo: https://github.com/yusufkaraaslan/Skill_Seekers - **skill-creator** (Meta, 159K★): Builds and optimizes new skills following Anthropic's own best practices - the AI agent researches a topic deeply and iterates on the skill until it holds up. · Why: The skill that makes all the other skills - we used it to build our client-specific ones. · Repo: https://github.com/anthropics/skills - **spec-miner** (Development): Point it at inherited, undocumented code - it reads everything, traces the data flows, and writes the spec that should have existed. - **stitch-skills** (Webdesign, 6.4K★): Google Stitch via MCP - generate full UI screens from a text prompt, piped directly into Claude Code. · Repo: https://github.com/google-labs-code/stitch-skills - **supabase-agent-skills** (Development, 2.3K★): Official Supabase skills - teach AI agents to work with Supabase databases, auth, storage, and edge functions from Claude Code. · Repo: https://github.com/supabase/agent-skills - **superpowers** (Development, 248K★): A complete software workflow for Claude Code: brainstorming, planning, subagent-driven development, TDD enforcement, code review, and git worktrees - with an isolated AI agent per task so long sessions do not drift. · Why: It is the backbone of how we build - every feature on this very website goes through its brainstorm and review steps. · Repo: https://github.com/obra/superpowers - **taste-skill** (Webdesign, 59K★): An anti-slop frontend framework with adjustable dials for layout variance, motion intensity, and visual density - soft, minimalist, brutalist, and redesign variants. · Repo: https://github.com/Leonxlnx/taste-skill - **the-fool** (Development): Challenges any big decision from five angles - devil's advocate, red team, pre-mortem - before you commit to it. - **theme-factory** (Webdesign, 159K★): Generates cohesive color palettes and typography systems you can apply across an entire product - maintained by Anthropic. · Repo: https://github.com/anthropics/skills - **ui-ux-pro-max** (Webdesign, 102K★): A searchable design intelligence database - 161 reasoning rules, 67 UI styles, 161 color palettes, 57 font pairings - that generates a complete design system before a line of code is written. · Repo: https://github.com/nextlevelbuilder/ui-ux-pro-max-skill - **video-toolkit** (Creative AI, 1.7K★): A full AI-native video production workspace - skills, slash commands, templates, and tools covering narration, scoring, rendering, and composition. · Repo: https://github.com/digitalsamba/claude-code-video-toolkit - **video-use** (Creative AI, 15K★): Agent-native video editing - cuts filler words, color grades, burns subtitles, and outputs a final MP4 from natural language instructions. · Repo: https://github.com/browser-use/video-use - **web-design-guidelines** (Webdesign, 29K★): Vercel's official skill that reviews UI code against their evolving web interface standards and returns actionable, file-level design, UX, and accessibility feedback. · Why: A second, stricter pair of eyes on layout and accessibility - complements impeccable well. · Repo: https://github.com/vercel-labs/agent-skills - **webapp-testing** (Development, 159K★): Anthropic's official web-app testing skill - opens a real browser via Playwright, runs interaction tests, captures screenshots, reports failures. · Repo: https://github.com/anthropics/skills ## Apps we use + links worth your time Full pages: https://agentmatik.ai/md/apps-we-use and https://agentmatik.ai/md/links ### AI agent platforms - **Claude Code**: Anthropic's agentic coding tool - terminal, IDE, and cloud. Why: Built this entire website and most of our client tooling. The most capable agent platform today. (https://claude.com/product/claude-code) - **Hermes**: An autonomous agent platform we run for specialized reporting work. Why: Part of our own AI agent fleet - reports land in Slack without anyone asking. (https://platform.futurehouse.org/) - **Manus**: Cloud AI agent that researches, browses, and produces finished deliverables. Why: Best for long, unattended research and document tasks - fire it off, come back to results. (https://manus.im) - **OpenClaw**: The open-source always-on AI assistant - runs 24/7 on your own hardware. Why: What we install for clients who want a personal AI assistant they fully own. (https://openclaw.ai) ### App stack - **Apify**: Web scraping platform with thousands of ready-made scrapers. Why: When a site has no API, Apify usually already has the scraper. (https://apify.com) · Deal: 30% off - **Apollo**: B2B lead database and outreach platform. Why: Lead lists that feed straight into automated outreach workflows. (https://apollo.io) · Deal: 50% off annual plans - ask us - **Claude**: Anthropic's AI - and Claude Code, the agentic coding tool this site is built with. Why: The single most-used tool in the company. Claude Code built and maintains this website. (https://claude.com) - **Composio**: The integration and auth layer for AI agents - manages OAuth for hundreds of tools so your AI agent never sees raw credentials. Why: The security middleware behind serious AI-assistant installs: scoped access, audit trail, instant revoke. (https://composio.dev) - **Context7**: Current documentation for any library or platform, served to AI agents over MCP - so they build against today's docs, not last year's training data. Why: Wired into our own Claude Code setup; it is why our builds do not use deprecated APIs. (https://context7.com) - **Cursor**: The AI code editor - for hands-on, editor-driven development. Why: When you want to stay in the driver's seat instead of delegating to an AI agent. (https://cursor.com) - **Duet Display**: Turns an iPad or spare screen into an extra display - including for a headless Mac mini. Why: The practical answer to running a Mac mini AI agent box without a monitor. (https://duetdisplay.com) · Deal: 20% off first purchase - **Figma**: Design and prototyping - where every interface starts. Why: Design source of truth; the MCP integration lets our AI agents read designs straight into code. (https://figma.com) - **Firecrawl**: Web scraping API built for AI - clean markdown from any page, search included. Why: Our AI agents' eyes on the web; powers the research behind this site's pages. (https://firecrawl.dev) · Deal: 50% off - **Fireflies**: Meeting recorder and transcription bot with solid webhooks and API. Why: The other half of our meeting pipelines - webhook-driven, automation-friendly. (https://fireflies.ai) - **Granola**: AI meeting notes that transcribe locally and summarize into your template. Why: Half of our meeting pipeline starts as a Granola transcript. (https://granola.ai) - **HeyGen**: AI video avatars - generate a presenting clone of yourself from one recording. Why: The fastest route to personal video content at scale. (https://heygen.com) · Deal: Partner deal - ask us - **Higgsfield**: Creative AI platform - image, video, and character consistency in one clean UI. Why: The best UI in creative AI right now; behind most visuals in our guides. (https://higgsfield.ai) · Deal: Partner deal - ask us - **Intercom**: Customer support platform with the strongest AI-agent support story. Why: Where support automation actually works in production. (https://intercom.com) · Deal: 12 months free (startup program) - **Kie.ai**: One API for photo and video generation models - Veo, Flux, Kling and more. Why: The cheapest way to call frontier image and video models from an automation. (https://kie.ai) · Deal: Credits deal - ask us - **Linear**: Issue tracking and product development - fast, opinionated, API-first. Why: Our product work lives here; AI agents file and update issues directly. (https://linear.app) · Deal: 1 year free - **ManyChat**: Chat marketing automation for Instagram, WhatsApp, and Messenger. Why: The engine behind comment-to-DM funnels - pairs perfectly with our tracked links. (https://manychat.com) · Deal: Up to 50% off - **Mixpanel**: Event analytics for products - strong funnels and retention reporting. Why: When the analytics questions get serious, this is what we reach for. (https://mixpanel.com) · Deal: Startup credits - **n8n**: The workflow engine for automation - visually orchestrate AI, humans, code, and APIs. Why: The backbone of nearly every automation we ship. If we could keep only one tool, this is it. (https://n8n.io) - **Notion**: Docs, databases, and the operating system for most of our client workflows. Why: Almost every automation we ship writes into Notion - meeting summaries, dashboards, trackers. (https://notion.com) · Deal: 3 months free - **PostHog**: Product analytics, session replay, and feature flags - open source. Why: One tool instead of four for understanding what users do in your product. (https://posthog.com) - **Shopify**: E-commerce platform for stores of any size. Why: The cleanest APIs in e-commerce - a joy to automate against. (https://shopify.com) - **Slack**: Team messaging - and the front door for AI agents and notifications. Why: Our automations report into Slack; it is where humans meet the workflows. (https://slack.com) · Deal: 30% off for a year - **Stripe**: The payment infrastructure - checkout, billing, invoicing. Why: Non-negotiable for taking money online; this site's Pro billing runs on it. (https://stripe.com) · Deal: Atlas 50% off - **Supabase**: Open-source Postgres backend: database, auth, storage, edge functions. Why: Our default backend for client apps - generous free tier and AI agents work with it well. (https://supabase.com) - **Superhuman**: The fastest email client - keyboard-first, with AI triage built in. Why: For inboxes that are not yet fully delegated to an AI assistant, this is the best manual cockpit. (https://superhuman.com) - **Vercel**: Deployment platform for modern web apps - push to git, it is live. Why: Zero-ops deployments; our default for client-facing apps. (https://vercel.com) - **Wispr Flow**: Voice dictation that works everywhere - speak, it types, in your style. Why: Most of Matt's long messages to his AI agents are dictated through it. (https://wisprflow.ai) ### Creators - **AI with Remy**: Hands-on AI tutorials with a practical, non-hype angle. Why: Consistently useful walkthroughs - rated Great in our creator database. (https://www.youtube.com/@aiwithremy) - **Claire Vo**: The How I AI podcast - practitioners showing exactly how they use AI at work. Why: Real screen-share workflows instead of talking heads. (https://www.youtube.com/@howiaipodcast) - **Greg Isenberg**: Startup ideas and AI playbooks, episode after episode - the channel behind The Startup Ideas Podcast. Why: Essential-tier: a steady source of automation ideas worth stealing. (https://www.youtube.com/@GregIsenberg) - **Matthew Berman**: AI news and tutorials covering OpenClaw, LLMs, and open-source AI. Why: The fastest way to stay current on the OpenClaw ecosystem. (https://www.youtube.com/@matthew_berman) - **Nate Herk**: AI automation tutorials on n8n and Claude Code - 528K subscribers and the AI Automation Society community. Why: Essential-tier in our book: the closest channel to how we actually build for clients. (https://www.youtube.com/@nateherk) - **Noe Varner**: Automation tutorials and the NoeAI community - strong on competitor research and ads workflows. Why: Several of our marketing skills trace back to his classroom material. (https://www.youtube.com/@noeautomates) - **Roman Knox**: AI content and design workflows, with a strong Gumroad library. Why: Rated Great in our database - especially for creative AI workflows. (https://www.instagram.com/roman.knox) ### Dashboard elements - **AdminLTE**: The most popular admin template of all time (Bootstrap). (https://adminlte.io/) - **Ant Design**: Dense enterprise tables and forms - the standard for ops-heavy admin UIs. (https://ant.design/) - **ApexCharts**: Interactive charts with 100+ ready samples to copy. (https://apexcharts.com/) - **CoreUI**: Admin components for Bootstrap, React, Vue, and Angular. (https://coreui.io/) - **Cruip Mosaic**: Free analytics dashboard template - React + Chart.js. (https://github.com/cruip/tailwind-dashboard-template) - **Cult UI**: The highest visual craft for animated and marketing components - shadcn-installable. (https://www.cult-ui.com/) - **Efferd** (our pick): The dashboard look we benchmark against - lift elements from here into Claude Code. (https://efferd.com/) - **Flowbite Admin**: Free Tailwind dashboard widgets from the Flowbite team. (https://github.com/themesberg/flowbite-admin-dashboard) - **HeroUI**: Gorgeous out of the box, built on React Aria for strong accessibility. (https://www.heroui.com/) - **Kokonut UI**: A large free pool of modern animated components - easy shadcn install. (https://kokonutui.com/) - **Mantine**: Broadest batteries-included React system - ~100 components plus hooks, dates, and charts. (https://mantine.dev/) - **Motion Primitives**: Reusable, tasteful animation building blocks rather than full sections. (https://motion-primitives.com/) - **MUI X**: Enterprise-grade data grid and pickers for dense data UIs. (https://mui.com/) - **Radix UI**: Headless, unstyled, gold-standard accessible primitives - the engine under shadcn. (https://www.radix-ui.com/) - **Recharts**: The default React charting library - composable and battle-tested. (https://recharts.org/) - **shadcn dashboard example**: The official shadcn dashboard example - closest free starting point to the Efferd look. (https://ui.shadcn.com/examples/dashboard) - **shadcnblocks**: 1,600+ ready page sections purpose-built for shadcn (one-time paid). (https://www.shadcnblocks.com/) - **Tabler**: Premium-quality free Bootstrap admin - plus 6,000 open-source icons. (https://tabler.io/) - **TailAdmin**: Admin template in 7 verticals - HTML, React, Vue, and Next flavors. (https://tailadmin.com/) - **Tailwind Plus**: Best-in-class marketing, app, and e-commerce blocks from the Tailwind team (paid). (https://tailwindcss.com/plus) - **Tremor**: 35+ dashboard blocks - KPI cards, charts, tables - built for React. (https://tremor.so/) - **Tremor OSS dashboard**: Open-source SaaS dashboard template on Next.js. (https://github.com/tremorlabs/template-dashboard-oss) - **v0**: Vercel's prompt-to-UI generator on a shadcn + Next stack - own and review the output. (https://v0.dev/) - **Windmill**: React + Tailwind dashboard with dark mode out of the box. (https://github.com/estevanmaito/windmill-dashboard) ### Design sources & MCPs - **21st.dev**: Community shadcn component marketplace - plus the Magic MCP so AI agents can pull components directly. (https://21st.dev/) - **Aceternity UI**: Stunning animated hero, 3D, and spotlight components - use sparingly, they are heavy. (https://ui.aceternity.com/) - **AstroWind**: The top Astro + Tailwind landing theme on GitHub. (https://github.com/onwidget/astrowind) - **Cruip free templates**: Free SaaS landing templates (Simple Light, Open) - solid full-page starts. (https://cruip.com/free-templates) - **daisyUI**: Semantic Tailwind component classes - works on any stack, even plain HTML. (https://daisyui.com/) - **Design Spells**: A catalog of delightful UI details - the little touches that make products feel alive. (https://designspells.com/) - **Fig Components**: A library of free Figma components to pull into design files. (https://figcomponents.com/) - **Flowbite**: 400+ Tailwind components with official React, Vue, Svelte, and Angular packages. (https://flowbite.com/) - **Flowrift**: 80+ vanilla HTML + Tailwind blocks - no framework required. (https://flowrift.com/) - **HyperUI**: The largest free copy-paste Tailwind component set - zero dependencies. (https://hyperui.dev/) - **Magic UI**: 150+ animated, shadcn-compatible components - marquees, beams, text effects. (https://magicui.design/) - **Mamba UI**: 150+ components you can copy as HTML, Vue, or JSX. (https://mambaui.com/) - **Meraki UI**: 144 clean Tailwind components with first-class RTL and dark mode. (https://merakiui.com/) - **Preline**: 220+ polished blocks and marketing sections - the best-looking free library. (https://preline.co/blocks/) - **React Bits**: 110+ animated React components, free and copy-paste. (https://reactbits.dev/) - **shadcn/ui**: The foundation: accessible copy-in primitives you own - the standard everything else builds on. (https://ui.shadcn.com/) - **ShaderGradient**: Animated gradient shaders you can configure visually and drop into a site. (https://shadergradient.co/) - **Tailblocks**: 60+ ready Tailwind blocks - stale but still a fast starting point. (https://tailblocks.cc/) - **TailGrids**: 100+ free blocks in HTML, React, and Vue flavors. (https://tailgrids.com/) ### Learning videos - **NVIDIA GTC keynote: the agentic era**: The keynote where NVIDIA's CEO declared agentic systems the new computer - the context behind this whole category. (https://www.youtube.com/watch?v=jw_o0xr8MWU) - **OpenClaw on a Hostinger VPS**: Step-by-step: deploy OpenClaw on a cheap VPS (KVM2, Ubuntu 24.04) - the exact path we use for cloud installs. Why: The tutorial we point clients to when they ask what hosting an AI assistant actually looks like. (https://www.youtube.com/watch?v=BhjK2Gr0Ryc) - **OpenClaw on Amazon EC2**: Deploying OpenClaw on an EC2 Ubuntu instance, end to end. Why: The AWS alternative to the VPS route, for teams already on Amazon. (https://www.youtube.com/watch?v=04wh2Hlgbds) - **Running local models with Claude Code**: How to run local models alongside Claude Code - when privacy or cost pushes you off the cloud APIs. Why: The practical walkthrough for the on-your-own-hardware question we get most. (https://www.youtube.com/watch?v=O2k_qwZA8HU) ### Podcasts & newsletters - **Behind the Craft**: Peter Yang's show on how the best builders actually work with AI - hands-on interviews, not punditry. Why: One of the few AI shows where guests share real workflows you can copy the same day. (https://www.youtube.com/@peteryang) - **Lenny's Newsletter & Podcast**: The product and growth newsletter - plus the interview podcast behind it. Why: The single best source on product, growth, and how real teams work. His Product Pass inspired this page's design. (https://www.lennysnewsletter.com/) - **The Startup Ideas Podcast**: Greg Isenberg's podcast - concrete startup ideas and AI playbooks, episode after episode. Why: Where a lot of our automation experiments start as someone's throwaway idea. (https://www.gregisenberg.com/) ### Prompt libraries - **Design Prompts**: 31+ design styles rendered from the same data, each with an AI-ready prompt to recreate the aesthetic. Why: Fastest way to calibrate a style: compare identical content across 31 aesthetics, then copy the winner. (https://designprompts.dev/) - **InstaLanding.ai**: One-shot prompts for cinematic, scroll-based, and interactive landing page components. Why: The one-shot components actually look premium out of v0 or Lovable - no ten-round prompt surgery. (https://instalanding.ai/) - **Jiro**: A design prompt library to vibe-code better websites - premium templates and components by niche (travel, e-commerce, SaaS). Why: The strongest art direction per niche - pick the vertical, get a coherent visual language. (https://jiro.build/) - **Lovable Prompts**: 101+ ready-to-use prompts for Lovable - SaaS apps, landing pages, UI/UX components, and backend logic. Why: The deepest Lovable-specific set - useful even as a checklist of what to specify. (https://lovable-prompts.com/) - **Motionsites.ai** (our pick): Premium prompts built specifically for animated, motion-based hero sections. (https://motionsites.ai/) - **shadcn space**: Shadcn UI blocks, templates, and components with a one-click Copy Prompt feature for v0, Lovable, and Bolt. Why: Pairs perfectly with our shadcn-first workflow - the prompts are pre-optimized for AI builders. (https://shadcnspace.com/) - **StyleKit**: Tailwind UI prompts - copy a prompt per component to generate Tailwind CSS designs. Why: Prompt-per-component granularity for Tailwind-native builds. (https://www.stylekit.top/en/tailwind-ui-prompts) - **Superdesign**: A free, open library of hundreds of UI and design prompts - dashboards, landing pages, components, animations - each with a live preview. Why: The biggest free prompt library we know: search the grid, preview the result, copy the prompt. (https://superdesign.dev/) - **TurnConcepts**: Copy-ready prompts and UI concepts for ChatGPT, Claude, and Cursor - prompt plus rendered UI proof in every card. Why: You see the rendered result before you copy - no prompt roulette. (https://turnconcepts.com/) - **ui-prompts (GitHub)**: An open-source collection of reusable prompts for consistent UI generation, including versioned dashboard prompts. Why: Versioned prompts you can diff, fork, and adapt - the engineering approach to prompt libraries. (https://github.com/arhamkhnz/ui-prompts) - **WebsitePrompts**: A free library of website prompts for ChatGPT, Claude, v0, Cursor, and Bolt - landing pages, SaaS sites, portfolios. Why: Tool-agnostic base library - a solid starting prompt for almost any site type. (https://websiteprompts.com/) ### Skills & repositories - **AI Templates**: 1,000+ Claude Code components: 842 skills, 421 agents, 281 commands, 86 MCPs, and 57 hooks. Why: Huge surface area - useful for discovering what categories of components even exist. (https://www.aitmpl.com/) - **anthropics/skills**: Anthropic's official open-source skill collection. Why: First stop for official, maintained skills - frontend design, skill creator, canvas, themes. (https://github.com/anthropics/skills) - **Awesome Claude Code**: Selectively curated list of the best skills, hooks, slash commands, orchestrators, and plugins for Claude Code. Why: Stricter curation than most awesome-lists - things here tend to actually work. (https://github.com/hesreallyhim/awesome-claude-code) - **Awesome Claude Skills**: Composio's curated list of Claude skills, resources, and tools. Why: The best-maintained general directory when you want breadth. (https://github.com/ComposioHQ/awesome-claude-skills) - **Awesome OpenClaw Skills**: The community directory of OpenClaw skills. Why: First stop when extending an OpenClaw install. (https://github.com/VoltAgent/awesome-openclaw-skills) - **Claude Skills Guide (jeffallan)**: A hand-written guide and catalog of practical Claude Code skills by category. Why: Where several of our development skills (Feature Forge, Code Reviewer, The Fool) come from. (https://jeffallan.github.io/claude-skills/skills-guide/) - **ClawHub**: The skill marketplace for OpenClaw - browse by downloads, install by name. Why: Where the OpenClaw entries on our Skills page come from. (https://clawhub.ai/skills?sort=downloads) - **gstack**: Garry Tan's exact Claude Code setup - 23 opinionated tools acting as CEO, designer, engineering manager, release manager, and QA. Why: A complete, working process for shipping with Claude Code - not a toolbox but a sprint: think, plan, build, review, test, ship. (https://github.com/garrytan/gstack) - **Marketing Skills**: CRO, copywriting, SEO, analytics, and growth engineering skills for Claude Code. Why: The copywriting, seo-audit, and content-strategy skills are the standouts. (https://github.com/coreyhaines31/marketingskills) - **skills.sh**: The leaderboard and registry for agent skills - installs, stars, and security audits per skill. Why: Our primary source for vetting skills before they touch a client machine. The layout inspired our own Skills page. (https://www.skills.sh/) - **SkillsMP**: Marketplace for OpenClaw and cross-platform agent skills. Why: The place to find OpenClaw-specific skills with a review layer. (https://skillsmp.com) ### Website inspiration - **Awwwards**: Award-winning websites - the high end of what the web can look like. (https://www.awwwards.com/directory/) - **Dribbble**: The classic design showcase - what designers are shipping right now. (https://dribbble.com/shots/popular) - **Framer Marketplace**: Production-ready Framer templates - useful as structure references even off-Framer. (https://www.framer.com/marketplace/) - **Godly**: Small, ruthlessly curated collection of high-end web design. (https://godly.website/) - **Land Book**: Curated gallery of real landing pages, searchable by industry and style. (https://land-book.com/) - **Lapa Ninja**: 7,300+ landing pages, tightly categorized - fastest way to see 20 takes on one pattern. (https://lapa.ninja/) - **Mobbin**: 400k+ real app screenshots - how the best products actually solve each screen. (https://mobbin.com/) - **One Page Love**: One-page sites and templates - the reference for single-page storytelling. (https://onepagelove.com/) - **SaaS Landing Page**: The best SaaS landing pages since 2020, organized by section and element. (https://saaslandingpage.com/) - **SiteInspire**: Clean, filterable showcase of web design - strong on typography-led sites. (https://www.siteinspire.com/) - **Trickle**: 50 prompts organized by visual style, mood, and interaction type. (https://trickle.so/) - **Vibe Code Components**: Premium React component prompts that generate design-ready UI. (https://vibecodecomponents.com/) --- This file is regenerated from the live site content. Canonical URLs: cite https://agentmatik.ai pages; markdown mirrors live under https://agentmatik.ai/md/.