The autonomous pipeline that takes ideas from signal to store. 32 AI agents. 5 stages. One decision-maker.
Every app flows through the same battle-tested pipeline. Each stage has dedicated agents, hard gates, and decision cards.
You are the Chief Decision Officer. Your only inputs are binary approve/reject decisions. The agents handle everything else.
Three ideation modes scan for opportunities: market-pull signals, technology-push from domain expertise, or clone-and-differentiate from existing apps.
Automated landing pages with Formspree forms and Plausible analytics. Multi-channel distribution. Conversion rate as the primary metric.
The Orchestrator coordinates spec, code, test, and review agents through sprint cycles. Compile checks, test execution, and UX walkthroughs are automated.
From signal scanning to App Store optimisation, the pipeline covers the full lifecycle.
8 sources: App Store reviews, Reddit, Google Trends, regulatory changes, competitor reviews, Twitter/X, Trustpilot, Product Hunt.
Two-pass evaluation: prosecution files a kill brief, defence responds with four checks. Only one idea promoted per batch.
iOS (Swift 6 / SwiftUI) and Web (TypeScript / React / Next.js). Platform set per-app with core + overlay agent composition.
Hard gate between build and monetisation. Sean Ellis 40% test, D7 retention, core loop engagement, NPS baseline.
After each TestFlight deploy, synthesise user feedback into feature health, amendment proposals, and backlog adjustments.
Action queue with dependency graph scheduling, parallel dispatch, cascade blocking, and full audit trail.
Shared APOSDesignSystem package with 15 components, semantic tokens, and per-app theming via environment injection.
Cross-app A/B test tracking. Results feed back into triage, specs, and PMF gate decisions for transferable insights.
7 research modules: market sizing, regulatory context, competitor analysis, user research, technical feasibility, monetisation, value chain.
Four categories: pipeline agents drive workflow forward, gate agents block on failure, advisory agents enrich without blocking, utility agents maintain the framework.
Coordinates build subagents, constructs context bundles, manages action queues. Never writes code directly.
Scans 8 signal sources, frames JTBD, scores ideas 1-5 with owner proximity and trend coupling modifiers.
Three-role adversarial team: Prosecutor, Defence, Judge. Genuine separation of concerns in independent contexts.
Writes code to satisfy numbered acceptance criteria. Platform-specific via core + overlay composition.
First Wow Moment, Hook Model, core loop, D2 retention, accessibility audit, persona-wow alignment.
Hard gate between Phase 2 and Phase 3. Four checks must pass before monetisation investment.
Flags 11 monetisation issue types. Enriches decisions without blocking the pipeline.
Detects framework changes via git diff, builds a sync plan, applies approved changes across all artifacts.
What changes when you stop stitching agents together manually.
APOS is Apache 2.0 licensed. Every agent definition, schema, and slash command is in the repo. Fork it, adapt it, make it yours.
Clone the repo. Run /generate-ideas. Let the pipeline do the rest.