Agent Skills

LaunchSaaS ships an agent skill with commands to scaffold, iterate, and customize apps in the monorepo

LaunchSaaS ships an agent skill — a set of commands that encode this monorepo's conventions (template structure, capability wiring, i18n, schema reuse) so the agent makes contextually correct edits instead of guessing.

The skill lives at .agents/skills/launchsaas/ and is the primary interface for building on LaunchSaaS with an AI agent.

How to Invoke

Call a command with /launchsaas <command> <args>:

/launchsaas create AI writing tool for marketers
/launchsaas new-feature user notifications with read-marking
/launchsaas doctor

Run /launchsaas with no command to print the command list.

Commands

Command	Args	Purpose
`create`	`<description> [--ref <url>]`	Copy the best-fit template into a new app + write a next-step checklist. Copy-only.
`new-page`	`<description>`	Add an App Router route page or UI surface. Use `content` for MDX content.
`new-feature`	`<description>`	Full-stack feature: DB table (if needed) + service + API + UI.
`rebrand`	`[app]`	Update name, tagline, logo/favicon/og assets, i18n strings, SEO, marketing homepage.
`add-capability`	`<type> [provider]`	Wire a capability (payment/email/storage/cache/newsletter/cron) into an app.
`auth`	`<subcommand> [args]`	Better Auth workflows: OAuth providers, captcha, magic link, email/password, 2FA, passkeys.
`integration`	`<type> [provider]`	Configure app integrations that are not capabilities or auth features (analytics/live-chat).
`content`	`<subcommand> [args]`	Add MDX content: blog posts, docs pages, changelog entries, static pages.
`add-email`	`<description>`	Add a transactional email template (enum + switch case + i18n fan-out).
`prune`	`<module> [app]`	Safely remove an unused template module. Destructive — needs a clean git tree.
`add-locale`	`<locale> [app]`	Add a language to an existing app.
`doctor`	`[app]`	Config health check: flag capabilities that resolve to `null`.

Examples

/launchsaas create AI tool directory --ref https://theresanaiforthat.com
/launchsaas new-page dashboard analytics page
/launchsaas rebrand launchsaas
/launchsaas add-capability email resend
/launchsaas add-capability cron qstash
/launchsaas auth add-provider google
/launchsaas auth add-captcha cloudflare-turnstile
/launchsaas integration analytics plausible
/launchsaas content add-page "privacy policy"
/launchsaas content add-doc "API rate limits"
/launchsaas add-email trial expiring reminder
/launchsaas prune blog
/launchsaas add-locale fr
/launchsaas doctor

Typical Lifecycle

Building a new product usually flows through the commands in this order:

create — scaffold from the closest template. This is copy-only; it hands you a checklist rather than a fully wired app.
Commit the scaffold — all follow-up commands check for a clean git tree. Commit now so they don't block you: git add apps/<slug> package.json && git commit -m "feat: scaffold <slug>"
rebrand — make it yours: name, logo, favicon, OG image, marketing homepage.
add-capability / auth / integration — turn on the providers and services you need (payments, email, social login, captcha, analytics, live chat). Each finishes with a setup checklist.
new-feature / new-page / content — build your product's unique UI surfaces and MDX content.
prune — drop template modules you don't use (e.g. blog, docs).
doctor — verify env, auth, and capability wiring before you ship.

create never edits src/capabilities.ts or src/env.ts. Capability wiring is delegated to add-capability so you never end up in a half-configured state where code references a provider whose env and dashboard resources don't exist.

Safety Model

Archive before mutating. Every command that edits existing files checks git status first and asks you to commit or stash if the tree is dirty.
prune is destructive and requires a clean tree (or explicit override), so a removal is always reversible with git stash pop.
Scoped writes. Commands stay inside the target app plus root workspace metadata; they won't silently change shared packages/* or other apps.
Docs stay in sync. Commands that touch a code surface in the Documentation Sync table also update its paired doc page.

Resources

AI-Assisted Development — how the agent docs are layered
Providers — capability wiring reference
Project Structure — the monorepo map