Slash Commands
The commands the kit adds to Claude Code, what each reads or writes, and what its answer looks like, from the sign-in check to the webhook verifier.
A slash command is a routine you type, and only you: every command in the kit is marked so that Claude never runs one on its own. Each one follows a skill step by step, reads before it writes, and confirms every change with you. They are Claude Code's; the skills behind them work in every client.
| Command | Argument | What it does |
|---|---|---|
/subscriby:connect | [oauth|token] | Sign Claude in to Subscriby through the bundled MCP server, or walk the personal-access-token route, then prove the connection with the account's details. |
/subscriby:docs | <question or topic> | Look a question up in the Subscriby documentation and answer from the page's own Markdown, with links. |
/subscriby:doctor | [project name or id] | Diagnose a project's connectors, places and recovery readiness, read-only, and report every gap with the tool or dashboard step that fixes it. |
/subscriby:launch | — | A guided launch from an empty project to a live one, following the launch-project skill step by step and confirming every write before it runs. |
/subscriby:mint-token | [purpose, e.g. read-only analytics, a webhook subscriber, a support assistant] | Propose the least set of abilities a personal access token needs for a purpose, and the steps to mint it, without ever handling the token itself. |
/subscriby:status | — | One card on the account's state, read-only. Account, team and plan, open alerts and notifications, the health of every installation, the inbox, Copilot Credits and recent activity. |
/subscriby:verify-webhook | [node|php|python] | Write a Subscriby webhook signature verifier and its test into the current repository, in the project's language and conventions, with the signed fixture that proves it. |
/subscriby:weekly | [project name or id] | The week in figures against the week before, read-only: revenue, payments, new and lost members, failed renewals and recoveries, trials ending, the inbox, the installations, the pass windows ahead. |
Typed, never automatic
Commands are the one thing in the kit that Claude cannot invoke by itself. Ask in plain words and the matching skill fires; type the command and the routine runs in its fixed order. Both reach the same tools with the same confirmations.
/subscriby:connect
Checks whether the bundled server is authenticated. When it is, the command reads your account through the server and answers with the account, the team the connection acts on, the plan's capabilities, the connected accounts and the alert destinations, so you know whose reach Claude has before anything else happens. When it is not, it walks the sign-in (oauth) or the personal-access-token route (token) and proves the result the same way. It reads; it changes nothing.
/subscriby:status
One card about the account: the team and its plan, open alerts and unread notifications, any installation whose health is degraded, Copilot Credits and the most recent activity. Read-only, and the first thing to run in a new session when something feels off.
/subscriby:doctor [project]
The project's diagnosis: each connector installation with its health, the connector doctor run against it, each gated place's standing, the recovery readiness checklist and the webhook endpoints with their recent failures. Every gap is reported with the tool or the dashboard page that fixes it. The doctor only reads; fixing is a separate, confirmed step.
/subscriby:launch
The guided launch from an empty project to a live one, in the order the launch skill prescribes: the project's fields, the connector connected by you, the places linked, the perks, a live payment method, plans created switched off and then published, the portal published, the go-live checklist. Each write is shown and confirmed before it runs, and the command stops at any step you decline.
/subscriby:weekly [project]
The week in figures against the week before, like the Copilot digest: revenue, new and churned subscribers, active count, the best-selling plan, failed payments and what recovered, with the period and the baseline named beside each number and the tool calls listed at the end. Read-only.
/subscriby:docs <question>
Finds the right documentation page from the site's llms.txt index, fetches that page's own Markdown twin, and answers from it with the links, so the answer carries what the page says today rather than what Claude remembers.
/subscriby:verify-webhook [language]
Writes a Subscriby webhook receiver into the repository you are working in, in the project's language and conventions: the signature verifier (raw body, constant-time comparison, a 300-second tolerance, a list of secrets for rotations), the route that deduplicates on the event id and answers quickly, and a test that proves four cases against a signed fixture (a valid delivery, a stale one, a tampered one, one signed with the next secret). It runs the test and shows the result, then tells you the three steps left: register the endpoint, store the secret, fire a test delivery.
/subscriby:mint-token [purpose]
Proposes the least set of abilities a personal access token needs for a purpose (read-only analytics, a webhook subscriber, a support assistant, an operator), from the ability catalogue, with the team and project scope and the dashboard steps to mint it. It never asks for the token and never suggests granting everything.
Related
- Skills for creators: the skills behind
launch,doctor,statusandweekly. - Skills for developers: the skills behind
verify-webhookandmint-token. - Safety: why a command still asks before a destructive tool runs.
How is this guide?
Last updated on
Skills for Connector Authors
The skills that walk Claude through building a connector on the SDK, from an empty package to a submission, with the manifest, the ports, the data rules, the fakes and the conformance kit at hand.
Subagents
The three subagents the kit adds to Claude Code, when Claude delegates to them on its own, how to ask for one, and the generated allowlists that keep them read-only.