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.
A subagent is a separate Claude with its own instructions and its own, smaller set of tools, started for one job and answering with a report. The kit ships three, each read-only by construction: the tools each may use are a generated allowlist of the server's read tools (plus file reading for the reviewer), so none of them can change anything however it is asked.
subscriby:community-analyst
Answers questions about a Subscriby community's figures and trends from the analytics and list tools, with the period, the baseline and the tool calls behind every number, never changing anything. Use proactively when the user asks how a project is doing, who lapsed, which plan sells, why revenue moved, or for a report.
subscriby:integration-reviewer
Reviews a repository's Subscriby integration (a webhook receiver, an API client, an automation) against the published contract and reports concrete findings with file and line, never editing anything. Use proactively after someone writes or changes code that calls api.subscriby.net or receives Subscriby webhooks, or when asked to review an integration.
subscriby:launch-auditor
Audits a Subscriby project against the go-live checklist from its live state and reports every gap with the fix, changing nothing. Use proactively before a creator announces a project, right after a launch, or when asked whether a project is ready to sell.
When they run
Claude delegates to a subagent on its own when the request matches the agent's description: a question about how a project is doing goes to the analyst, a request to review a webhook receiver to the reviewer, a check before an announcement to the auditor. You can also ask for one by name, for example "use the subscriby:launch-auditor agent on Studio Lab", and Claude starts it with your words as its task.
Each agent answers in a fixed shape so a report reads the same every time: the auditor as a checklist of gaps with the fix for each, the analyst as figures with the period, the baseline and the tool calls behind them, the reviewer as a table of findings with file and line, severity and the fix.
Read-only, by the allowlist
An agent's tools are listed in its definition and regenerated from the catalogue with the rest of the kit: every tool marked read-only on the server, and nothing marked additive or destructive. The hook that guards destructive tools never has to fire for an agent, because an agent cannot reach one.
What a report looks like
Related
- Slash commands:
/subscriby:doctorand/subscriby:weeklyare the same reads as a routine you type. - Safety: the guard, the ability model and the audit trail.
How is this guide?
Last updated on
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.
Permissions and Safety
The guard plugin whose hook makes every destructive tool ask, what destructive means on the Subscriby server, least-privilege tokens, the audit trail, members' data, and the settings a team admin uses.