dx-lifecycle-center — скилл для Claude Code
Скилл dx-lifecycle-center для Claude Code и других платформ, который управляет соединениями с организациями Salesforce и разворачивает метаданные. Запускается при командах подключения сред или переноса компонентов.
Apache-2.0 110 установок ★ 1 183 обновлён 2026-10-23
npx skills add forcedotcom/sf-skills --skill dx-lifecycle-center -a claude-codeЧто делает dx-lifecycle-center
Этот скилл автоматизирует работу с целевыми средами в Salesforce. Основная часть его функций выполняется из исходной организации, где запущен DX Inspector. Инструмент позволяет выводить список подключённых сред, регистрировать новые песочницы или рабочие окружения, а также выполнять деплой, валидацию и быстрый перенос компонентов между средами.
Внутри реализована логика разбора запросов, определяющая тип операции: от получения статуса текущих соединений до запуска процесса развёртывания. Скилл ожидает подтверждения пользователя перед изменением состояния сред, если выполняется деплой или подключение. Для корректной работы требуются установленные в системе jq версии 1.6 или выше и sf версии 2.67.0 или новее.
Какие задачи решает
- вывести список доступных для деплоя сред
- подключить новую песочницу или рабочую среду
- валидировать метаданные перед отправкой
- развернуть компоненты в целевую организацию
Примеры запросов агенту
- Покажи список всех подключённых сред
- Подключи мою песочницу к текущему проекту
- Разверни компоненты в целевую среду sandbox
Как установить dx-lifecycle-center
1Через Skills CLI (Vercel)
npx skills add forcedotcom/sf-skills --skill dx-lifecycle-center -a claude-code2Вручную: положить файлы
~/.claude/skills/dx-lifecycle-center/SKILL.mdПапка скилла целиком; для одного проекта — .claude/skills/dx-lifecycle-center/
1Через Skills CLI (Vercel)
npx skills add forcedotcom/sf-skills --skill dx-lifecycle-center -a codex2Вручную: положить файлы
.agents/skills/dx-lifecycle-center/SKILL.mdДля всех проектов — ~/.agents/skills/
1Через Skills CLI (Vercel)
npx skills add forcedotcom/sf-skills --skill dx-lifecycle-center -a cursor2Вручную: положить файлы
.cursor/skills/dx-lifecycle-center/SKILL.mdCursor читает и .agents/skills/, и .claude/skills/
1GitHub CLI
gh skill install forcedotcom/sf-skills skills/dx-lifecycle-center2Вручную: положить файлы
.github/skills/dx-lifecycle-center/SKILL.md1Через Skills CLI (Vercel)
npx skills add forcedotcom/sf-skills --skill dx-lifecycle-center -a gemini-cli2Командой платформы
gemini skills install https://github.com/forcedotcom/sf-skills --path skills/dx-lifecycle-centerВручную: положить файлы
.devin/skills/dx-lifecycle-center/SKILL.mdСтарый путь .windsurf/skills/ тоже читается
Через Skills CLI (Vercel)
npx skills add forcedotcom/sf-skills --skill dx-lifecycle-center -a opencodeClaude.ai
Упакуйте папку в zip и загрузите в Settings → Capabilities → Skills
По шагам, с командами для каждой платформы: Как установить скилл в Claude Code: пять способов
Вопросы о dx-lifecycle-center
Как установить dx-lifecycle-center в Claude Code?
Выполните npx skills add forcedotcom/sf-skills --skill dx-lifecycle-center -a claude-code в терминале проекта.
Что делает dx-lifecycle-center?
Он помогает управлять доступом к организациям Salesforce и автоматизирует развёртывание метаданных между ними прямо из чата с агентом.
Что нужно для работы dx-lifecycle-center?
В системе должны быть установлены утилиты jq (>=1.6) и sf (>=2.67.0).
Исходный текст
SKILL.md · Apache-2.0
---
name: dx-lifecycle-center
description: "Lifecycle Center for DX Inspector and Builder Central. Four jobs: (A) list the orgs connected to DX Inspector; (B) connect a new org (sandbox or scratch, production, OrgFarm) as a deploy target; (C) connect or check the Lifecycle Center org, also written LifecycleCenter, lifecycle-center, ALMDevOpsCenterHub, or Hub org; (D) deploy, validate, or quick-deploy components that already exist in the user's org to a connected org, and poll a simple-deploy requestId. TRIGGER on: show, list, or pick connected orgs or environments; connect, add, log in to, or repair an org; connect the Lifecycle Center org; deploy, validate, push, or publish components (metadata) to another org, URL, org id, alias, or environment name. Short asks like \"connect a new org\" or \"deploy to my sandbox\" match. NOT for promoting work items (dx-devops-promote), local source deploys (platform-metadata-deploy), or publishing sites or packages. Deploys move components between orgs, not local files."
metadata:
version: "1.0"
domains: ["Developer Experience"]
minApiVersion: "67.0"
relatedSkills:
- "dx-devops-promote"
- "platform-metadata-deploy"
cliTools:
- tool: ["jq"]
semver: ">=1.6"
- tool: ["sf"]
semver: ">=2.67.0"
---
# Lifecycle Center: Connect Orgs and Deploy Metadata
This skill manages which orgs a user can deploy to, and deploys to them. Everything runs from the **source org**: the org the user built in, where DX Inspector runs. All connections and environments are stored there.
## Operations
| Op | Name | What it does | Details |
|---|---|---|---|
| **A** | List orgs | Shows the orgs connected to DX Inspector (the deploy targets) and whether each is ready in this workspace, or the current org's details. Also resolves a target the user names. | `references/list-orgs.md` |
| **B** | Connect an org | Connects a new sandbox or scratch, production, or OrgFarm org as a deploy target, or repairs a half-finished connection. The user signs in through a link. | `references/connect-org.md` |
| **C** | Lifecycle Center org | Connects, repairs, or checks the Lifecycle Center org (credential `ALMDevOpsCenterHub`). | `references/hub-org.md` |
| **D** | Deploy | Deploys, validates, or quick-deploys components from the source org to a connected target, then polls to the end. | `references/deploy.md` |
Connect (B) and deploy or quick deploy (D) change an org, so they wait for the user's confirmation first. Listing, status checks, validate runs, and polling do not change a target org.
## Key terms
| Term | Meaning |
|---|---|
| Source org | The org the user built in. Use the alias they name, else the default org from `sf org display --json`. Components must already exist in it. |
| DX Inspector environment | A `DevopsEnvironment` record in the source org with a Named Credential. The deploy targets are exactly these records. |
| Deploy target | An environment in state `selectable`. An `expired` or `unauthenticated` environment is not a target. |
| Lifecycle Center org | The Lifecycle Center hub. A separate connection that is never a deploy target; it is shown in a list only when connected, labelled Lifecycle Center (Production Org). |
## Routing (deterministic)
Scripts and references are paths inside this skill's folder (the folder that holds this file). This skill owns every request below. Pick the operation by these rules **in order; the first match wins**. Do not hand these requests to another skill, and do not answer them from general knowledge.
| # | The request | Operation |
|---|---|---|
| 1 | Names the Lifecycle Center org, in any request, however it is spelled: "Lifecycle Center", "lifecyclecenter", "lifecycle-center", "lifecycle_center" (matching rules in `references/hub-org.md`) | **C** |
| 2 | Deploy, validate, quick deploy, push, or publish components (metadata) to an org, URL, org id, alias, or environment name; or gives a `requestId` or validation id | **D** |
| 3 | Show, list, or pick connected orgs or environments; "which orgs can I deploy to"; the current org's details | **A** |
| 4 | Connect, add, log in to, reconnect, or repair an org or its credential; **sync**, **refresh**, or **re-sync** an org's workspace access; or a `sf` command on a connected org failed with `Session expired or invalid` / `INVALID_SESSION_ID` | **B** (the sync procedure, `references/connect-org.md`) |
| 5 | Anything else about connected orgs or DX Inspector environments | **A** |
The Lifecycle Center org rule (1) has two exclusions: "Dev Hub" is never the Lifecycle Center org, and plural phrases such as "DevOps Center orgs", "DX Inspector orgs", "environments", or "orgs connected to DX Inspector" are not the Lifecycle Center org; route them by rules 2 to 5. Details: `references/hub-org.md`.
**Reading the wording.** Before applying the rules, normalize the request: lowercase it, treat hyphens, underscores, and extra spaces as single spaces, and ignore punctuation, one-letter typos in words of five or more letters, and filler such as "please", "can you", "I want to". (Rule 1 also uses the two forms described in `references/hub-org.md`.) Then match on any word in the groups below; the exact phrase does not matter.
| Operation | Action words (any form: "deploy", "deploying", "deployment") | Object words |
|---|---|---|
| **D** | deploy, deployment, validate, validation, check only, dry run, quick deploy, quickdeploy, push, send, ship, move, copy, release, publish, add or install (when the object is a component, never a site, article or package version) | components, metadata, classes, flows, objects, or a named org as the destination ("to", "into", "onto") |
| **B** | connect, connection, add, link, hook up, hookup, authorize, authenticate, log in, login, sign in, set up, setup, register, onboard, reconnect, repair, fix, refresh | org, environment, env, sandbox, scratch, production, prod, orgfarm, org farm |
| **A** | show, list, see, view, display, which, what, pick, choose, available | orgs, environments, envs, targets, connections, "where can I deploy" |
A request with a **D** action word and a destination, or that names a component and a destination whatever the verb ("add AccountService to my QA sandbox"), is **D**, even when it also contains a **B** word. A **B** action word whose object is an org or environment makes the request **B**, even when words like "connected" or "linked" also appear ("reconnect my sandbox, it shows not connected"). A request with only **B** words is **B**. A request with only **A** words, or a question about what is connected, is **A**. A bare `requestId` or "status of my deploy" is **D**. If no group matches, use the vague-prompt table below.
A request that mixes jobs (for example "connect my sandbox and deploy AccountService to it") takes the lowest-numbered match, here **D**, which runs **B** itself when the target is not connected.
**Vague prompts.** If the message is only "deploy", "push my changes", "publish my changes", "connect an org", or similar, do not guess and do not call any API. Ask exactly one question that fixes the first missing input:
| Intent | Ask, in this order |
|---|---|
| Deploy | 1. Which components? (`references/deploy.md`, Step 1) 2. Which target org? |
| Connect | 1. "Is it a sandbox or scratch org, a production org, or an OrgFarm org?" 2. What short name should the environment have? |
| Not clear which | "Do you want to connect an org or deploy components?" |
A bare "deploy" with no DX Inspector context is a question, not a deploy.
## Rules that always apply
- Targets come from the DX Inspector environments in the source org, never from `sf org list` or local aliases. (A connect also registers the environment name as a workspace alias for the user's convenience; that does not make aliases a source of targets.)
- Never deploy to a target that is not `selectable`, never to the source org, and never to the Lifecycle Center org.
- The Lifecycle Center org appears in a list only when it is connected, as one labelled row, Lifecycle Center (Production Org), with no button, never a deploy target and never in Sync all. When it is not connected the list does not mention it.
- Confirm before every call that changes an org: connect (`connect-org.md`, Step 1) and deploy (`deploy.md`, Step 4). `validate` is the only mode that needs no confirmation. A click on a Sync / Refresh button, or an explicit ask to sync, is the confirmation for a workspace sync; listing never syncs by itself.
- A timeout (`TARGET_ORG_AUTHENTICATION_TIMEOUT`) is transient: retry once after about 10 seconds yourself, then stop. Only a 401 or 403 means the sign-in is stale.
- When a connect started from a deploy and finishes, tell the user the org is connected and that you are continuing the deploy, then ask for the deploy confirmation. Connecting never deploys by itself.
- A sandbox and a scratch org are the same type (`SANDBOX`). Treat their URLs alike.
- **Always double-quote an org alias in a shell command.** Environment names may contain spaces (`Dev 12 Org`), so write `--target-org "Dev 12 Org"`, never `--target-org Dev 12 Org`. Quote every alias, name or id taken from user input or from the resolver output, in every `sf` command and script argument.
- **Fix what can be fixed yourself.** An error a retry or a sync can clear (expired session, timeout, a missing workspace alias on an org the task needs) is handled in place: sync or retry, then carry on with the task and say what you did in one line. Do not hand the user a button or a question for it. Ask only when the fix needs them: a browser sign-in, a name clash, or an org-changing action they have not asked for.
- Make one attempt per API call. Never retry in a loop, resubmit after a timeout, or invent an id, credential name, or status.
- Never ask for, print, or pass an access token. Never show row ids or credential names.
- Never say a browser was opened. Give the login link and wait for the user to reply **done**.
- Do not guess a missing input. Ask one question for the first missing input.
- Ask the user to confirm an org type that the URL host does not decide. Never infer it from an environment name.
## What the user sees
| After | Reply with |
|---|---|
| **A** list | The table of usable orgs, the action buttons (Sync / Re-sync / Refresh, plus Sync all N when two or more need it), the one left-out line when needed, and the closing sentence (`list-orgs.md`, section 4). No raw JSON, no status column, no other offers. |
| **B** sync (button click or failed `sf` command) | One line per org (synced, timed out, sign-in needed, name clash, or not available), then the list again. No sign-in link for a timeout (`connect-org.md`, *Sync workspace access*). |
| **A** not enabled | The one-sentence message in `list-orgs.md`, section 2. Nothing else. |
| **B** connect | The sign-in layout with a button-style link (`connect-org.md`, Step 2), then after **done** the connected layout: environment name, org host, type, and status (`connect-org.md`, Step 5). |
| **C** | The exact sentences in `hub-org.md`. When connecting, the same sign-in and connected layouts as B, in hub form. |
| **D** | A plain-language report: what was deployed, from where, to where, and the outcome (`deploy.md`, Step 7). No raw JSON. If the deploy is still running after about 2 minutes: the `requestId`, a note that it is in progress, and a request to send the id back to get the status (`deploy.md`, Step 6). |
## If something fails
| Failure | Where to look |
|---|---|
| The list or target lookup fails | `references/list-orgs.md`, section 2 |
| A connect step fails | `references/troubleshooting.md` |
| A deploy or validate fails | `references/error-handling.md` |
| The target login is stale | Sync it first (`references/connect-org.md`, *Sync workspace access*) and retry the step yourself. Only a 401/403 that a sync cannot fix needs the user's sign-in (repair path), and after it you retry without asking |
| A `sf` command on a connected org says the session expired | Sync that org's workspace access (`references/connect-org.md`, *Sync workspace access*), then retry the command once |
## References
- `references/list-orgs.md`: Operation A. The resolver script, match statuses, the list table, and the not-enabled handling.
- `references/connect-org.md`: Operation B. Modes, Steps 0 to 5, and rules for connecting an org.
- `references/hub-org.md`: Operation C. The Lifecycle Center org: hub names, replies, and rules.
- `references/deploy.md`: Operation D. Steps 1 to 7 for deploy, validate, quick deploy, and polling.
- `references/request-body.md`: deploy, validate, and quick-deploy request bodies and test levels.
- `references/connect-api.md`: Environment API and `setupNamedCredential` request and response shapes, status checks.
- `references/troubleshooting.md`: connect error codes and recovery.
- `references/error-handling.md`: deploy error codes.
- `examples/workflows.md`: worked prompts for each operation.
- `scripts/resolve-environment.sh`: read-only list and identifier resolver.
- `scripts/hub-get.sh`: read-only GET against the Lifecycle Center org through its credential (`references/hub-org.md`).
- `scripts/poll-deploy-status.sh`: read-only status poller. Reads the Simple Deploy endpoint and `DevopsRequestInfo`, and reports component errors on failure.
© forcedotcom · Apache-2.0 · forcedotcom/sf-skills