What problem does it solve? Teams often commit to building features that later fail because of hidden third-party limits, platform policy walls, or architecture assumptions that don't survive contact with the real codebase. This Skill runs an adversarial pre-planning discovery workflow that tries to break a feature idea cheaply, before any build plan is written. ## Core Features & Use Cases - Parallel integration research: Dispatches one research agent per named third-party service (Stripe, RevenueCat, IAP, Firebase, etc.) to surface hard limits, quotas, policies, pricing, and known gotchas with cited sources. - Codebase reality check: Explores the scoped areas of a monorepo-aware codebase to find what exists, what conventions apply, and what would push back against the new feature. - Adversarial three-lens panel: Attacks a draft approach along constraints & policy, edge cases & failure modes, and alternatives & prior art, classifying findings as blockers, significant, minor, or false positives. - Decision artifact: Produces a .discovery/<slug>/discovery.html report with weighed architecture options, open questions, and a recommended next step (plan, pivot, or don't build). - Use Case: Before building creator subscriptions with Stripe Connect and RevenueCat, run discovery to uncover SKU limits, App Store policy constraints, and failure modes like mid-flow payment failures, then hand the surviving approach to plan-phases. ## Quick Start Ask the assistant to run feature discovery on your feature idea, for example: "Pressure-test my creator subscriptions feature that uses Stripe Connect and RevenueCat before we plan it."
