Skip to content

Ship an enterprise app

GrowthOS is the planning system of record. Cursor (or Claude) is the implementation engine. Humans keep three gates: PRD quality, UX approval, and per-ticket review. Fast depth is for spikes. Production is the enterprise bar: module-sharded PRD, uncapped UX, engineering blueprint, feature map, NFR tickets, and a Sprint 0 platform pack (CI/CD, environments, auth, observability, backups, load-test harness, security).

Path A is a new idea (/projects/new → Idea Lab, or /product/idea-lab?depth=production). Path B is a shipping product (/projects/new?mode=existing&depth=production). Later waves use Continue planning on the same project. Coding agents work Native tickets over MCP — one task per turn.

  • Who: Planner shipping a product you will operate — not a one-week spike
  • Time: Days, not 15 minutes — Production uses 3–8× more AI runs
  • Plan: Pro+ for sprints and unlimited MCP handoffs. Native board; Connected is Enterprise-gated.
  • Toggle Production before generate so later jobs inherit depthMode on the PRD.
  • Hit Build-ready (readiness 80+) before the Production backlog pipeline.
  • Link Figma frames onto UX Spec screens, then implement from figmaUrl.
  • Run the sprint loop in the application repo: enrich → code → tests → Agent handoff → DONE.
  • You will operate this product in production — auth, CI, observability, and real UI matter.
  • You want Path A and Path B documented as one enterprise operating model.
  • Planner seat. Pro+ if you will start an ACTIVE sprint this week.
  • For UI: Figma MCP plus a gos_ key. For agents: MCP in the product repo, not only GrowthOS itself.
  • Chooser enterprise links: /projects/new → Idea Lab · Production, or Bring existing · Production.
  • Direct: /product/idea-lab?depth=production and /projects/new?mode=existing&depth=production.
  • Keys: Settings → API Keys & Agents. Sprint board → Connect coding agent.
  1. Pick Path A (new idea) or Path B (shipping product). On generate, leave Production selected — Fast is not the enterprise bar.
  2. Path B only: paste docs or run growthos-product-context (seed_product_context → upsert_living_prd) for this wave — never a rewrite backlog.
  3. Write a sharp brief (users, non-goals, compliance, integrations). Generate the PRD. Run Gap Scan, share for review, and do not finalize until readiness is 80+ (Build-ready). Production backlog generation refuses a weaker PRD.
  4. Approve delivery phases (required). Approve UX Spec for any UI product — skip only for APIs/jobs with no screens.
  5. Connect GrowthOS MCP + Figma MCP. Run growthos-prototype-design: critique visualDirection, draw P0 then P1 frames, write figmaUrl back. Do not ship cream+serif, dark+acid-green, or Inter-on-white shadcn as the whole identity.
  6. Review the engineering blueprint and feature map (Production rail). Build the backlog and confirm Sprint 0 / platform-pack stories (CI, auth, observability) sit ahead of feature work.
  7. AI sprint plan → start ACTIVE. Leave the next ticket unassigned (or assigned to you). In the product repo, mint a personal gos_ key and say: Run the GrowthOS sprint loop for project proj_…
  8. Confirm before claim. enrich_task before any code. Honor fidelity: production (no mock gateways). Run the test plan including a negative/authz case. Verify every AC, post ## Agent handoff, then DONE. First claim should be a Sprint 0 ticket (CI or auth), not a random feature.
  9. Say continue for the next task. Ask before advance_sprint. New scope: Continue planning on this project (run UX Spec in Idea Lab first if the wave has new screens). If briefs show stale_as_is, re-seed product context.
  • The PRD is Production depth, Build-ready (80+), and finalized. UX Spec P0/P1 have figmaUrl when the wave has UI.
  • Backlog includes Sprint 0 platform-pack work. An ACTIVE sprint has a claimable head ticket.
  • Each DONE ticket has verified AC, passing tests, and an Agent handoff. Commits include the ticket code (TASK-421).
  • One gos_ key per human — never a shared company key. Agents must not overwrite Connected synced status.
  • If brief.gaps is non-empty (thin AC, missing REQs, stale As-is), ask — do not invent. Implement UI from Figma, not generic Tailwind.
  • Production uses the same monthly AI-run quota — no extra SKU. Budget runs before you generate a second Production PRD.
  • Leaving Fast on for a product you will operate.
  • Skipping delivery phases, skipping UX Spec on a UI product, or overriding the readiness-80 gate to “just get a backlog.”
  • Asking the agent to finish the whole sprint, rewrite an existing product, or create a second project for wave 2.