Build process. How the node's development survives the end of a session, and how a step is planned, carried, proven and closed — in the core, a template workshop or an element. Load it at the start of every session of development; the moment the owner gives any task, however small; when a substep is finished; when the context runs out; when a step or a group of steps closes; before creating any folder of your own to track work; and every time something happens in the work: the owner decides, an expensive fact is measured, a commit is made, a defect is found on the way, a long operation is about to start. The rule that changes behaviour: nothing is compacted — the state is written to current-steps.md in words, at once, and the session restarts clean.
Когда и как подключается СВОЙ домен этого узла, и чем это отличается от временного адреса. Читать, когда человек говорит «хочу свой домен», «адрес всё время меняется», «ссылка перестала работать», «сайт не открывается снаружи»; когда собираешься обратиться к Cloudflare через MCP; и прежде чем что-либо менять в туннелях, DNS или адресе сайта. Правило, меняющее поведение: у узла ДВА пути в интернет, и они не взаимозаменяемы — временный адрес живёт часами и не несёт авторизации, постоянный принадлежит человеку.
Pages whose DATA is dynamic — a user's own screen, an admin list, anything read at request time. Load it before building a page under `(protectedLayer)`, an `/api/*` door that serves protected data, or when the owner says "a page listing users / orders / their own bookings", "show what this person has", "an admin table". The thing you cannot guess: in this project such a page is still PRERENDERED — a static shell with dynamic holes — and the two lines that would make it honestly dynamic (`cookies()`, `headers()`) take the whole layer down with them. The second thing: a door that proxies another service can never be more permissive than that service.
Waves of agents instead of one — what a dynamic workflow is, what it may be pointed at, and what it costs. On the node it runs only in a session the person confirmed as ultracode in Telegram (skill ultracode-session, which sends you here by path). Read it there before the first wave, or when the owner asks how workflows work. The thing you cannot see from inside a session: the agents a workflow spawns run on THIS machine with THIS session's permissions, and they edit files auto-approved — so the only thing standing between a fan-out and a large, tidy, expensive wrong answer is a command that judges each unit without you.
Where a task about elements talking, agents, deals, prices, ratings, Lagrix or exchanging elements belongs: i2i (inside the node), a2a (work for outside agents through Lagrix) or f2f (whole elements exchanged with other Fractera projects through Lagrix). Load it before touching the element menu (`lib/web4-items/item-tree.ts`: «Build» with its groups «Inside the node (i2i)» and «Exchange with Fractera (f2f)», the section «A2A»), the Deals page, the Lagrix rating card, a deal summary, a connector between elements, an F2F card, or anything that mentions payment between agents, a token, escrow, a wallet or reputation; and whenever a person asks «where does this go?». The rule that changes behaviour: inside the node aim at a connector over the API so that no model is needed; to the outside a Fractera element is an agent like any other — no advantage, no handicap; the element itself changing hands is f2f.
LINKS — what this project does differently about them. Load it when you write body text with links, add a button or card that points somewhere, or the owner says "link these pages", "add a source", "mention their site". Short on purpose: link attributes, anchor text and interlinking practice you already know. Two things you cannot know are here — an outgoing link is the architect's decision, not yours, and this engine has its own written form for internal links, with a gate behind it.
Addresses, routes and delivery order — the project's own map service. Load it when the owner says "show it on a map", "how far is it", "in what order should the courier drive", "find the address", or before you reach for Google Maps and a paid key. The thing nobody guesses from inside the guest layer: this exists, it costs nothing per call, and it answers with real road distances — but only for the region whose map was loaded, and loading one is a heavy background job that the owner starts, not you.
Moving an existing project onto this architecture: the launch GATE that must pass before anything begins, nineteen stages in order with what each ends with, and ten probes to run before writing code over somebody else's data. Load it when the mode is `migration`, when the owner says "we are moving this project" / "port my app" / "here is the repo we migrate from" — and **at the very first message**, because §1 decides where the session must be standing, which cannot be repaired later. Load it again at the start of every session of such a move: a migration outlives dozens of them and you locate yourself by its stage. The mode itself, the four questions about rights, and the rules for foreign code live in `CLAUDE.md` §`migration`; this file does not repeat them.
LANGUAGES of this project — which set is authoritative, when you translate and when you deliberately do not, and where the debt for that is recorded. Load it before writing any user-visible string, any language cell, any dictionary, and whenever you wonder "how many languages does this need". Two things here you cannot infer from the code: the set belongs to the OWNER of this project and never to the number your neighbours happen to have, and during development you write in ONE language on purpose — the missing translations are a recorded debt, closed by a separate run at the end.
The media store — where uploaded images, video and icons actually live. Load it before adding an upload, before showing a picture the owner uploaded rather than one committed to the repository, when an image answers 401 or renders blank, and before writing image resizing of your own. Two facts decide almost everything and neither is visible from the code you can see: a stored file is referenced BY NAME, never by id, because ids differ on every server — and the picture reaches the visitor through this app's own route, never from the data host, which is why a secret is involved in showing a public image at all.
How a page is added in a Fractera WEB4 ITEM — as a FOLDER OF DATA read by one route template, never as a new `page.tsx`. Load it before creating anything under `app/`, when the owner asks for "a new page", "a section", "an article", "a tab with N pages", "the same page in every language", and whenever a build is slow and you are about to look for the cause. The thing you cannot guess: every route file is compiled and type-checked on its own with everything it imports, so a site built as one `page.tsx` per page gets slower with every page and ends at many minutes per build — measured here as 1252 s for 300 page files against 98 s for the same 300 pages through one template. The build succeeds either way; nothing warns you until the owner waits an hour.
How this project holds data about PEOPLE — names, contacts, documents, health, anything that identifies someone — and what may cross the border to a model provider. Load it BEFORE designing any table that will hold a person, any support or medical feature, any AI call whose input comes from a user's own words. Also load it when the owner says "the data may not leave the country", "this is medical", "GDPR", "personal data", "we may not send that to OpenAI". The rule the whole skill turns on: the model receives the CASE, never the PERSON — the two are separate tables, and they are joined again only on the owner's own server, after the answer comes back.
Everything about PLATFORM-CONFIG — the eleven switches that decide which capabilities the site offers: top menu, footer pages, cookie banner, sign-in, breadcrumbs, FAQ, theme toggle, width toggle, language switcher, socials row, offline cache. Use when the owner says "turn off the menu", "remove the cookie notice", "hide the login button", "we don't need breadcrumbs", "hide the social icons", "add a header", "the switch does nothing", or anything about a capability being present or absent. ALSO use BEFORE building any header, footer, consent strip, theme or language control of your own — they already exist, they are switched, and a hand-rolled copy answers to nobody.
The UI primitives — one owner per kind of thing, and gates that refuse hand-rolled copies. Load it before writing a heading, a modal, a side panel, a button, an empty state, or anything that "just needs a div"; also when a gate refuses your markup and the message names a primitive you have not met. The point of the layer: shadcn/ui with ONE owner per kind of thing, and every colour taken as a TOKEN — so the owner repaints his whole project from the panel in a single move. An element with a literal colour does not break, it silently leaves that system. There is exactly one modal here, a hand-built one is refused by a gate, headings never carry raw classes, and a size never shrinks as the screen grows. Each rule replaced a real mess, not a preference.
The product dossier — the fourth config, and the only one that decides WHAT you are allowed to build. Load it in `cases` mode before writing any code, when the owner says "add a product" / "this is for the shop" / "why is that page here", and whenever you are about to name a folder after a product. Two things here are counter-intuitive and both are load-bearing: the product's `id` means nothing and never changes, while everything you would naturally use as an identifier — the name, the address, the type — is something the owner edits freely; and what a product HAS built is never stored, only counted, because a stored list of files disagrees with the files.
The INSTALLABLE side of the site — per-language manifest, icons, iOS splash screens, the service worker and the install prompt. Load it when the owner says "make it installable", "add it to the home screen", "the icon is wrong", "it opens white when I launch it", "add offline", or when you touch `app/manifest.ts`, `app/[lang]/manifest.webmanifest`, `public/sw.js`, the register island, or the icon and splash generators. Two defects here are invisible from code: an app that cannot be installed at all because the manifest has no icons, and — the expensive one — a service worker serving yesterday's page, which also makes every measurement you take in the browser a lie.
ROLES — who sees what, and where that is actually decided. Load it before building anything behind authentication, before adding a role check, when the owner says "only managers should see this" or "let administrators do it too", and whenever a screen works for you and refuses somebody else. Two things you cannot guess: the visible gate on a page is a sign rather than a lock, and a door that fronts another service may never be more permissive than that service — get that backwards and the interface promises access, then refuses it, with the cause split across two places.
One address, or one address per record — the decision hidden inside almost every request about "a page for X", and the conversation that must happen before you build either. Load it whenever the owner asks for "a client page", "a page for each post", "a card for a user", "show one order", when he asks why there are numbers or words in an address, and — most important — whenever anything is being REMOVED from an address that already has a parameter in it. He will almost never use the words "dynamic route": that is the whole problem, and recognising the request behind his words is your job, not his. Not to be confused with `use-dynamic-pages`, which is about dynamic DATA on one fixed address; this is about the ADDRESS itself.
WHERE a route goes on disk and who may open it — the two layers, the four permission groups, and the folder shape a page must have. Load it before creating any folder under `app/`, when the owner asks for "a page for managers / for buyers / for me only", when a gate refuses a folder you just made, or when you are about to import something from a neighbouring group. The mistakes here are not stylistic: a page in the wrong layer has to be moved rather than configured, a folder one level too high turns the whole public layer into a tab and fails gates on code you never touched, and a borrowed import across permission groups breaks nothing today and everything later.
The SECTION layer — the closed catalogue of section kinds every page is built from, and what it takes to add one. Load it when a page needs a look the catalogue does not have, when you are about to write anything under `sections/`, or when the owner says "make a section that…". The decision this skill exists for is made before the first file: a new kind is a promise the platform keeps forever, and most of what feels like a new kind is a widget belonging to one route. Adding one wrongly does not fail loudly — it quietly makes the project's look un-maintainable.
A folder with a SKILL.md file: a name, a description of when to use it, and instructions. Claude loads a skill only when the task matches its description.
How do I use one I find here?
Copy the folder into your project's .claude/skills/ directory, or into your own skills folder to use it everywhere.
What do the warnings mean?
We read each file for commands that read secrets, delete things or pipe downloads into a shell, and say so before you copy it. No warning is not a promise that a file is safe.
Which skills worked for people?
Open a skill to see its discussion. Reports from people and their agents are coming.