What problem does it solve? Writing PRDs that both humans understand and AI agents can implement directly is hard — most PRDs are either too vague for execution or too prescriptive for collaboration. This Skill produces structured PRDs with enough technical detail (schema, APIs, architecture) for Claude Code to implement without ambiguity, while keeping the problem framing readable for stakeholders. ## Core Features & Use Cases - Four entry modes: Guided interview, Context Dump with automatic gap analysis, Best Guess with [ASSUMIDO] markers, and From Brief inheriting artifacts from upstream skills like idea-to-brief and deep-research. - Three complexity levels: Lean, Standard, and Comprehensive templates with mandatory sections like Out of Scope, 3-layer guardrail metrics, Epic Hypotheses, and Testing Decisions with prior art. - Validation scorecard: Evaluates any PRD across 5 weighted dimensions (Problem, Scope, Requirements, Technical, Risks) with a 7.5 delivery threshold and automatic anti-pattern correction. - Task generation: Converts approved PRDs into vertical slices (tracer bullets) labeled Direct or Blocking, each demonstrable end-to-end. - Use Case: You paste a rough idea for a multi-tenant SaaS dashboard; the Skill detects complexity, asks only about gaps, and delivers a Standard PRD with RLS schema, API contracts, and ordered implementation tasks. ## Quick Start Ask the assistant to create a PRD for your project idea, or paste your existing context and say "faz o PRD" to trigger automatic mode detection.
