agentleFS

Claude skills and agent skills

Skills real projects publish on GitHub, most starred first. Each one says what it will make an agent do before you copy it.

Worked for most · soon
FracteraSkill

shape-the-need

Fractera/Agent-Engineering-Infrastructure/.claude/skills/shape-the-need/SKILL.md

Work process. The node's method for any goal, wish, need or trouble a person brings, in any words, and for what the node can be for them — the forces the architecture is built on, the closed set of shapes (an answer, a route in an element, a new element, a separate boundary of control, the network) and the method that tests every shape on the node's facts and gives the variants that survive, with a pick and one question. Load it with any skill of a field, never instead of it.

682d agoReviewMountCreate a free account to mount
FracteraSkill

skill-creator

Fractera/Agent-Engineering-Infrastructure/.claude/skills/skill-creator/SKILL.md

Create new skills, modify and improve existing skills, and measure skill performance. Use when users want to create a skill from scratch, edit, or optimize an existing skill, run evals to test a skill, benchmark skill performance with variance analysis, or optimize a skill's description for better triggering accuracy.

188 projects use a version of this

Often used withtaskflow-inbox-triagecommunity-led-growthecosystem-reference

682d agoReviewMountCreate a free account to mount
FracteraSkill

study-for-element

Fractera/Agent-Engineering-Infrastructure/.claude/skills/study-for-element/SKILL.md

Build process. The person wants to know how a finished product, site or SDK works, to build something like it — the element that will build it studies it in its own study pass; you only hand the request over. Load it when the person asks to study, take apart or reverse-engineer a site, an app or an SDK, asks «how is X made», or wants a product «like X».

682d agoReviewMountCreate a free account to mount
FracteraSkill

supervise-element

Fractera/Agent-Engineering-Infrastructure/.claude/skills/supervise-element/SKILL.md

Build process. Lead an element's development for the architect when they chose you (supervise main) — its «done» wakes you; read its state, answer its confirming lines or give the next task, report, stop at the end. Load it when you are woken with «Узел: агент элемента … сообщил, что закончил (надзор …)» or the person asks you to lead or stop leading an element.

682d agoReviewMountCreate a free account to mount
FracteraSkill

translate-pending

Fractera/Agent-Engineering-Infrastructure/.claude/skills/translate-pending/SKILL.md

Перевести всё, что записано долгом в development-docs/TRANSLATION-DEBT.md, и вычеркнуть строки той же правкой. Вызывать, когда владелец говорит «переведи что накопилось», «закрой долги по переводам», «добавь язык X везде», или когда он ВКЛЮЧИЛ новый язык в NEXT_PUBLIC_SUPPORTED_LANGUAGES и сайт на нём говорит по-английски. Переводит САМ агент по подписке — внешний сервис не зовётся и ключ не нужен. Не вызывать для одной строки в одном файле: это обычная правка.

682d agoReviewMountCreate a free account to mount
FracteraSkill

ultracode-session

Fractera/Agent-Engineering-Infrastructure/.claude/skills/ultracode-session/SKILL.md

Build process. A session the person turned to ultracode in Telegram (Opus → Ultracode → the warning → «Продолжить»): the one place where the node allows dynamic workflows and subagents, and how it ends. Load it only when this session's system prompt says «Ultracode — this session only» or Claude Code's reminder says ultracode is on — never to switch it on yourself.

682d agoReviewMountCreate a free account to mount
FracteraSkill

use-agentic-rag

Fractera/Agent-Engineering-Infrastructure/.claude/skills/use-agentic-rag/SKILL.md

The knowledge graph — documents in, written answers out. Load it when the owner wants the site to answer questions from company documents, when an answer must be composed across many files, before loading anything large into it, and when the base answers "unavailable" or answers nothing useful. Two things decide everything here and neither is visible from the code: loading is expensive while asking is cheap, so what you load is a budget decision; and the graph is built in the BACKGROUND, so a question asked right after loading honestly sees nothing.

682d agoReviewMountCreate a free account to mount
FracteraReads credentialsSkill

use-ai-generation

Fractera/Agent-Engineering-Infrastructure/.claude/skills/use-ai-generation/SKILL.md

WHEN this project calls a model, and what our law says louder than the AI SDK's own skill. Load it before writing anything that generates, extracts, classifies, summarises or decides with a model; before adding a tool the model may call; before touching one of the twelve files that still call `v1/chat/completions` by hand; and the moment the `ai-sdk` skill offers the AI Gateway, a model identifier, or a provider install. The foreign skill owns HOW the SDK works; this one owns WHEN it is used here, where the key comes from, and which of its suggestions are forbidden in this project. Four collisions, each resolved in our favour with the reason, so the foreign skill can be followed everywhere else with a clear head.

682d agoReviewMountCreate a free account to mount
FracteraSkill

use-aio

Fractera/Agent-Engineering-Infrastructure/.claude/skills/use-aio/SKILL.md

The MACHINE surface of the site — the markdown twin of every public page, `llms.txt` and `llms-full.txt`. Load it whenever you add a public section or page, move an address, wonder what a model sees when it visits, or the owner says "make it readable for AI", "add llms.txt", "ChatGPT quotes the wrong price", "the agent gets a 404". Same reader problem as search, opposite shape: a crawler sends a human to a page, a model comes itself and needs TEXT — page markup is half chrome to it. The trap this area produces is silent: a new section ships with no twin, `llms.txt` never learns it, and the only symptom is a model that answers about your site from something older.

682d agoReviewMountCreate a free account to mount
FracteraSkill

use-app-config

Fractera/Agent-Engineering-Infrastructure/.claude/skills/use-app-config/SKILL.md

Everything about APP-CONFIG — what the application IS: name, description, brand, logo and images, icons and PWA, author, SEO and OpenGraph, structured data, analytics, currency, the `nav` branch, and the open list of social networks (`seo.socialLinks`). Use when the owner says "change the site name", "put my logo in", "add my Instagram", "the footer shows the wrong Telegram", "set the description for search", "our Twitter is stale", or anything about the site's identity or its accounts. ALSO use BEFORE hard-coding any name, address, handle, brand string or profile URL into a component — every one of them already has a home here, and a hard-coded copy is invisible to the panel and to the owner.

682d agoReviewMountCreate a free account to mount
FracteraReads credentialsSkill

use-auth-providers

Fractera/Agent-Engineering-Infrastructure/.claude/skills/use-auth-providers/SKILL.md

HOW a person gets in — passwords, Google, an emailed link — and which of those the owner switches on himself without a line of code. Load it when the owner says "let people sign in with Google", "send a login link by email", "I forgot my password", or when a sign-in method is simply missing from the form. What you cannot see from the guest layer: the sign-in methods are decided by which KEYS exist in the environment of a service you do not own, so a missing button is almost never a bug in the code — and one method everybody assumes is there does not exist at all.

682d agoReviewMountCreate a free account to mount
FracteraSkill

use-auth

Fractera/Agent-Engineering-Infrastructure/.claude/skills/use-auth/SKILL.md

AUTHENTICATION and the four tables it owns — who the people of this project are, where they are stored, and what you may and may not do with them. Load it before any page or door that lists accounts, changes roles, shows "who am I", or touches `users`, `sessions`, `accounts`, `verification_tokens`. Two facts here are the opposite of what the architecture suggests: those tables live in the SAME database as your products (not in a separate auth database), and they are not yours — the auth service on :3001 writes them, and deleting any of the four destroys every account on the server.

682d agoReviewMountCreate a free account to mount
FracteraSkill

use-browser

Fractera/Agent-Engineering-Infrastructure/.claude/skills/use-browser/SKILL.md

The agent's EYES — `npm run eyes`, a browser with no window that ships with every node: it opens a page of this node, clicks, takes a screenshot and tells what lies on top of what. Load it before you show any built surface to a person (a page, a drawer, a banner, a menu), when a proof needs the eye rather than the wire — console errors, a layout that is formally correct and looks wrong, an element covered by another — and before believing any measurement taken through automation. Needs no extension and no account of the person. Acting in SOMEONE ELSE's interface (a registrar, Google, Resend) is `help-in-browser`, not this.

682d agoReviewMountCreate a free account to mount
FracteraReads credentialsSkill

use-channels

Fractera/Agent-Engineering-Infrastructure/.claude/skills/use-channels/SKILL.md

The project's outside conversations — today Telegram. Load it when the owner wants a bot, a notification "to my phone", a voice assistant in a messenger, or a way for people to ask the site questions from Telegram; and before you install a Telegram library or poll a bot from a route of your own. Two things decide everything and neither is visible from the code you can see: Telegram hands each update to exactly ONE reader, so a second poller does not add a feature, it silently eats half the messages — and that one reader is the channel service on the loopback, which sends, receives, transcribes voice and pushes every message into your own /api/telegram/hook. The bot is set up inside this project, at /{lang}/architect/telegram.

682d agoReviewMountCreate a free account to mount
FracteraSkill

use-chat

Fractera/Agent-Engineering-Infrastructure/.claude/skills/use-chat/SKILL.md

WHEN the vendored `ai-elements` skill is the answer, and the rules of ours that overrule it. Load it before you build anything that looks like a conversation — a chat with a person, a chat with a model, a feed of messages, a log of what a bot heard, an inbox. There is exactly one right answer in this project and it is never hand-built markup: the chat tool in `_tools/chat/`, built on AI Elements. Also load it before adding any `@ai-elements/*` component, because the CLI asks to overwrite our own primitives and a yes there silently changes the design system of the whole project.

682d agoReviewMountCreate a free account to mount
FracteraSkill

use-code-shape

Fractera/Agent-Engineering-Infrastructure/.claude/skills/use-code-shape/SKILL.md

The SHAPE the code of this project must keep, and the eighteen build-time validators that enforce it instead of trusting anyone's memory. Load it before writing a route, an API door, a component that is growing, a database table, or anything that moves on screen — and whenever a gate refuses your work and you are about to "work around" it. The rules here are not style: each one is a limit that, when crossed, produces a defect nothing else reports — a page silently recomputed on every request, an API door open to guests, a table that exists on one machine, a crawler served an empty box. The central law: NO page in this project is dynamic, including the ones whose data is.

682d agoReviewMountCreate a free account to mount
FracteraSkill

use-database

Fractera/Agent-Engineering-Infrastructure/.claude/skills/use-database/SKILL.md

Tables, and the two doors that lead to them. Load it before adding a table, before writing a query anywhere, when rows exist on your laptop but not on the server (or the other way round), and before reaching for raw SQL. The two facts nobody guesses: a table is declared in ONE constant in this repository and appears on both machines by itself, and the same data has a second door that runs arbitrary SQL — which is deliberately harder to open than the first, because it bypasses the protection the first one gives you for free.

682d agoReviewMountCreate a free account to mount
FracteraReads credentialsSkill

use-data

Fractera/Agent-Engineering-Infrastructure/.claude/skills/use-data/SKILL.md

The ONE door out of your layer — the data service — and the three levels of generality behind it. Load it before saying "the app cannot reach that", before inventing storage of your own, and whenever you need the database, the media library, vectors, the knowledge base, maps or channels. What you cannot guess from inside the guest layer: the platform's code is not in your repository, so the surface is invisible — yet it is far wider than the helpers you can see. Arbitrary SQL and full pass-through to internal services already exist; the door describes itself; and a request that is refused tells you which of the two permission classes you fell outside of.

682d agoReviewMountCreate a free account to mount
FracteraSkill

use-deep-memory

Fractera/Agent-Engineering-Infrastructure/.claude/skills/use-deep-memory/SKILL.md

Work skill. The person wants something kept for later or brought back — «запомни», «запиши», «сохрани», «не забудь», «напомни о», «что я говорил про…», /remember or /remind from the bot's menu, in any language and any words — or sends a photo, a document or a voice note with such words in the same or the next message. Load it before npm run remember or npm run remind.

682d agoReviewMountCreate a free account to mount
FracteraReads credentialsSkill

use-deploy

Fractera/Agent-Engineering-Infrastructure/.claude/skills/use-deploy/SKILL.md

How a change reaches the running site and how you prove it did. Load it before saying "deployed", "live", "check it now"; when a page looks stale or blank after a reload; when a build "passed" but nothing changed; and when you are about to write your own ssh/scp instead of the project transport. The rule that changes behaviour: a build log proves nothing — the deployment is proven by the commit in /api/health matching your hash AND the port being held by a pm2 descendant whose uptime grows. Also: when a change needs no deploy at all, and the four traps that have cost time here.

682d agoReviewMountCreate a free account to mount
CLAUDE.md vs AGENTS.md

About skills

What is a Claude skill?

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.