projected-start
/projected-start
Prowadzi od pomysłu biznesowego do działającej, przetestowanej usługi backendowej w projected, z rusztowaniem i regułami dla agentów.
- Autor
- martinzokov ↗
- Repozytorium
- projected ↗
- Licencja
- MIT
- Ostatnia zmiana repozytorium
- 31.10.2026
Ocena i bezpieczeństwo
- Pliki w paczce
- 1
- Skrypty
- 0
- Komentarze HTML
- 0
- Adresy URL w SKILL.md
- 1
Brak sygnałów. Repozytorium zmienione 31.10.2026, bez odwołań do wycofanych modeli i API.
MIT, otwarta: można używać i zmieniać, z zachowaniem informacji o autorze.
Co robi i kiedy
Prowadzi od pomysłu biznesowego do działającej, przetestowanej usługi backendowej w projected, z rusztowaniem i regułami dla agentów.
Oryginalny opis (EN)
Use when the user wants a new backend service, API or microservice built with projected, or asks to start, create, scaffold or kickstart a projected project, and there is no projected app yet (no package.json depending on @projected/*). Takes the user from a business idea to a running, tested, agent-ready service: picks the flavour, scaffolds it with the published CLI, installs, verifies, sets up the project's own skills and guard hooks, and builds the first feature.
Jak działa
- 1. Understand the service before creating it
- 2. Check the tools
- 3. Create the service
- 4. Install, generate, verify
- 5. Set the project up for agents
- 6. Build the first feature
- 7. Run it and hand over
- When something fails
Podgląd SKILL.md
--- name: projected-start description: "Use when the user wants a new backend service, API or microservice built with projected, or asks to start, create, scaffold or kickstart a projected project, and there is no projected app yet (no package.json depending on @projected/*). Takes the user from a business idea to a running, tested, agent-ready service: picks the flavour, scaffolds it with the published CLI, installs, verifies, sets up the project's own skills and guard hooks, and builds the first feature." category: software-development --- # Starting a projected service projected is a TypeScript framework for CQRS services. Its CLI, `@projected/cli` on npm, writes a complete service: layers, composition root, HTTP server, auth, persistence, and tests that pass from the start. Your job is to turn what the user wants into the right first command, check the result, and hand over a service they can grow. **Already in a projected app?** If a `package.json` here (or in a parent directory) depends on `@projected/*` and there is a `projected.json`, stop: this skill is not for you. Use the `projected` skill to build features in it. ## 1. Understand the service before creating it Ask only for what you cannot infer, in one message. You need four things: | You need | How to get it | Rules | |---|---|---| | the **service name** | the business capability it owns: `billing`, `enrolment`, `inventory` | kebab-case; it becomes the directory name and the API's name | | its **first aggregate** | the main thing it keeps consistent: an `invoice`, a `student`, an `account` | camelCase (`invoice`, `lineItem`) and a noun, not a table or a verb | | the **flavour** | see below; default to `state-based` | fixed for the life of the service | | the **first feature** | one thing a user of the service does: "issue an invoice" | a business operation, not CRUD | Choose the flavour from what the user says: - **`state-based`** (the default): each aggregate is stored as its current state, and the events its changes emit go out through an outbox. Pick it unless something below applies. - **`event-sourced`**: each aggregate is the sum of its events, and every change is kept forever. Pick it when the user needs an audit trail, a history of every change, "what did it look like on date X", or other services to replay what happened: ledgers, payments, compliance. If the choice is unclear, say which one you picked and why, in a sentence, and go on. Don't ask about anything the framework already decides: HTTP, database, auth, validation, OpenAPI and messaging all come with it. ## 2. Check the tools ```sh node --version # needs v22.12 or newer pnpm --version # needs pnpm 10 or 11 git --version ``` - Node too old or missing: ask the user to install Node 22 LTS or newer (nvm, fnm, Volta or nodejs.org). Don't install it for them. - pnpm missing: `corepack enable pnpm` (ships with Node), or `npm install -g pnpm` if corepack is unavailable. ## 3. Create the service From the directory the service should live in: ```sh npx --yes @projected/cli@latest new service <name> --flavour <state-based|event-sourced> --aggregate <aggregate> ``` The CLI refuses names its rules would reject, and writes nothing when it does. Its message says what to change: read it, fix the name, and run the command again. If the current directory is a pnpm workspace (it has a `pnpm-workspace.yaml`, here or above), create the service where the workspace's `packages:` globs will include it, and run `pnpm install` from the workspace root in step 4. Otherwise the service is a workspace of its own. ## 4. Install, generate, verify ```sh cd <name> git init pnpm install pnpm openapi && pnpm lock && pnpm catalog # openapi.json, contracts.lock.json, docs/catalog.md pnpm check # compile, tests, and projected check ``` `pnpm check` must pass before you go on. A fresh service passes it; if it doesn't, see **When something fails** below rather than changing the generated code to get green. ## 5. Set the project up for agents ```sh pnpm exec projected agents install git add -A && git commit -m "Create <name> with projected" ``` This writes `AGENTS.md`, the projected skills (in `.agents/skills/`, `.claude/skills/` and `.cursor/rules/`), each harness's hooks, and a git pre-commit hook that runs `projected guard --staged`. Commit afterwards, because `projected guard` checks what changed since the last commit. Your harness may load the new skills and hooks only in a new session. Don't wait for that: read `.agents/skills/projected/SKILL.md` now, and run `pnpm exec projected guard` yourself after each edit until the hooks are active. ## 6. Build the first feature Follow the golden path in `.agents/skills/projected/SKILL.md`. In short, start with the generator, then make its stubs say what the business says: ```sh pnpm exec projected new command issue-invoice --aggregate invoice --event InvoiceIssued ``` Names are business language: commands are `verb-noun` in kebab-case (`issue-invoice`, never `create-invoice` or `update-invoice`); events are past-tense PascalCase facts (`InvoiceIssued`). The generator refuses CRUD verbs and says why. Fill in the domain rule, the contract's input, output and errors, and the tests. Then: ```sh pnpm openapi && pnpm lock && pnpm catalog pnpm exec projected guard --full # done only when this is green ``` ## 7. Run it and hand over ```sh pnpm build && pnpm start # http://localhost:3000, in memory curl -s localhost:3000/health/ready curl -s localhost:3000/openapi.json ``` Every operation needs a signed-in caller by default. Sign-up is at `/api/auth/sign-up/email` (better-auth). With `DATABASE_URL=postgres://…` the service stores everything in PostgreSQL and creates its tables at start. Stop the server, then tell the user: - what you created: the service, its flavour and why, and the first feature with its endpoint (`POST /commands/<type>/version/1`); - how to run it and where its API is described (`openapi.json`, `docs/catalog.md`); - how to grow it: ask for the next feature in business terms, and an agent in this project will follow the installed skills. ## When something fails | Symptom | Cause and fix | |---|---| | `npx` gets `404 Not Found` for `@projected/cli` | the registry is unreachable, or a private registry in `.npmrc` does not proxy npmjs.org. Check `npm view @projected/cli version`. | | `ERR_PNPM_IGNORED_BUILDS` naming esbuild | pnpm 11 wants build scripts approved. Services made by current CLI versions allow esbuild in `pnpm-workspace.yaml`; in your own workspace, add `allowBuilds:` with `esbuild: true` to its root `pnpm-workspace.yaml`. | | `new: … exists and is not empty` | choose another name, or ask the user before touching what is there. | | `pnpm check` fails on a fresh service | don't edit generated code to get green. Run `pnpm exec projected check` for the precise rule, and report its output to the user with the version from `pnpm list @projected/cli`. | | `agents: not a git repository — no pre-commit hook` | run `git init`, then `pnpm exec projected agents install` again. | Never switch off a check to get green: no `eslint-disable`, `@ts-ignore`, `@ts-expect-error` or skipped test. The checks are what keep the service's design intact.
Wersje
W bazie nie ma innych skilli o nazwie projected-start.
Migawka bazy z 13.11.2026 · wersja paczki c6b8d635fc14 (skrót SHA-256 całej paczki)
Pobrania w ostatnich 30 dniach: 0 · ranking
Zgłoś problem z tym skillem: szkodliwa treść, błędny opis albo prośba autora o usunięcie.
Podobne skille
- 100projected
Prowadzi od wymagania biznesowego do działającej, udokumentowanej usługi w projected: od zmiany w kodzie po endpoint i testy. - 100projected-integrations
Pomaga połączyć aplikację projected z zewnętrznymi systemami: projekcjami, innymi usługami, API stron trzecich, brokerami i reakcjami na zdarzenia. - 95backend-standards
Zakłada nowy serwis backendowy w NestJS, TypeScript, PostgreSQL i Redis zgodnie ze standardami zespołu od pierwszego commita. - 90new-project
Zakłada nowe repozytorium projektu z prostego pomysłu: stos, testy, CI, zasady GitHub, pliki AGENTS.md i CLAUDE.md oraz specjalistów. - 100kickstart-scaffold
Generuje działający nowy projekt w wybranym stosie: szkielet aplikacji, CI, testy, bezpieczeństwo, bazę i wdrożenie. - 90create-project
Zakłada nowy projekt Dart lub Flutter z szablonów Very Good CLI i instaluje zależności, dobierając szablon do celu.