Ship product

Skill DobroslavRadosavljevic/skills/skills/ship-product

Personal AI agent skills I use in development workflows.

Install

npx -y skills add DobroslavRadosavljevic/skills --skill ship-product

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

1 stars1 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its author says it does

Copied from the file, not written here

Manually build a paid TanStack Start fullstack product end-to-end (need-based stack, Polar, auth or guest checkout, env gate before live testing). HUMAN INVOCATION ONLY — do not auto-apply. Use only when the user explicitly invokes ship-product / $ship-product.

SKILL.md

4.4 KB, as published. Nobody here has run it

Ship Product

Build one concrete paid product end-to-end. Not a sellable boilerplate kit.

Invocation

Manual / human-only. Apply this skill only when a human explicitly invokes it (for example $ship-product, "use ship-product", or "follow ship-product"). Do not load or follow it automatically because the task looks like a greenfield app, SaaS, or TanStack Start project.

Hard rules

Need-based installs — thin backbone first; add packages only when the product map requires them. See references/stack.md.
Env gate — implement features, UI, and integrations in code first. Do not run live provider calls, browser E2E against paid/AI/email flows, or claim the app works until a user-provided .env exists and validates. Details: references/env-gate.md.
Self-contained — this skill embeds its own stack, naming, and workflow rules. Do not depend on other agent skills.
No competitor source — do not copy ShipFast, Supastarter, or unrelated product repos.
Prefer bun / bunx in all command examples.

Workflow

Follow references/build-phases.md. Summary:

Brief — fill the product map (references/product-brief.md): identity Mode A or B, money, core loop, optionals.
Scaffold — TanStack CLI in the opened workspace (see workspace rule below). Docker Compose, t3-env, .env.example, README.md, AGENTS.md.
Foundation — Drizzle, UI kit, evlog, health. Mode A: Better Auth + Redis secondary storage. Mode B: guest order/access tables.
Domain — real product feature. TanStack Query for API; Legend State for local UI only.
Money + email — Polar + Resend wired against env schema (no live sends yet).
Shell + optionals — marketing, legal placeholders, then AI/PDF/R2/etc. only if needed.
GATE — stop; present .env.example checklist; wait for real .env.
Verify + fix — after .env validates: compose up, migrate, seed if useful, browser paid journey, fix until clean (references/verification.md).

Workspace rule

Situation	Action
Empty / new workspace cwd	Scaffold at workspace root
Non-empty workspace	Ask once; recommended default = product-named subfolder
Unrelated dirty files	Do not clobber or rewrite unrelated work

Identity modes

Pick one primary mode (details in references/product-brief.md):

Mode A — Accounts + dashboard: Better Auth, email verification required, optional Google OAuth, signed-in shell.
Mode B — Guest checkout: configure → email → pay → deliver via email and/or unique opaque URL and/or R2 download. No dashboard unless the brief adds it.

Architecture snapshot

One app, one process: TanStack Start + in-process Elysia at /api/*.
Feature-oriented layout; short path naming (references/naming.md).
Legend State = local UI only; TanStack Query = API/server state (references/architecture.md).
Redis when rate limits or Better Auth (secondary storage). Omit Redis for guest apps with neither.
Payments: references/payments.md.

Docs every product ships

README.md — humans: what it is, setup, local defaults, scripts.
AGENTS.md — agents: stack, layout, commands, env testing gate, naming, non-goals. Use references/agents-md.md.
.env.example — complete variable list with Compose defaults (references/env-gate.md).

Done before env gate

Product features + UI for the chosen identity mode
Integrations coded against env schema (not live-called)
Minimal marketing + legal placeholders
README + AGENTS.md + .env.example
Compose Postgres; Redis only if auth and/or rate limits
Typecheck / unit tests without cloud secrets preferred
Live browser / provider journeys: not yet

Stop condition (after .env)

Primary paid user journey works in a real browser for the chosen mode. Fix defects until none remain that block that journey.