What problem does it solve? Turning a feature request or spec into an executable, reviewable implementation plan is error-prone: requirements get missed, tasks are oversized, and quality gates are configured inconsistently. This Skill enforces a disciplined planning protocol that produces traceable, right-sized plans with persisted QA configuration. ## Core Features & Use Cases - Spec-gated planning with traceability: Checks for .swarm/spec.md, detects stale specs, and verifies every FR-### maps to at least one task while flagging gold-plating tasks with no requirement source. - Four-stage clarification funnel: Inventories uncertainties, classifies them (self-resolved, critic-resolved, research-needed, user-decision), consults a critic sounding board, and surfaces a structured decision packet before saving the plan. - Task granularity and phase rules: Enforces small/medium/large sizing, splits oversized tasks, deduplicates redundant test tasks, and mandates multi-phase structure for larger plans. - QA gate and execution profile persistence: Collects user choices for eleven QA gates, parallel coder count, and commit policy, then persists them via set_qa_gates and a lockable execution_profile in save_plan. - Use Case: A user pastes an external plan and asks to prepare it for implementation. The Skill offers spec generation or direct ingestion, runs the clarification funnel, saves a phased plan, collects QA gate selections, and hands off to critic review before execution. ## Quick Start Ask the agent to create an implementation plan for your feature, for example: plan the user authentication feature from my spec and configure the QA gates.
