prd-v07-epic-scoping_skill

Overview

This skill transforms v0.6 technical specifications into context-window-sized EPICs for v0.7 Build Execution. It creates EPIC entries with objectives, ID references, dependencies, branch conventions, and context budget guidance. The output is designed to feed v0.7 Test Planning and agent-driven implementation sessions.

How this skill works

On request (e.g., “create epics”, “scope work”, “what to build first?”) the skill inventories API-, DBT-, FEA-, and ARC- items, identifies natural boundaries, and sizes work into cognitive-boundary EPICs. It produces EPIC- entries that include context capsules (pre-load context, working room, session goals), explicit dependencies, branch names, and phase-based execution plans sized for AI sessions. The skill also sequences EPICs by dependency and infrastructure priority to create a DAG-ready build order.

When to use it

• When converting v0.6 technical specs into actionable v0.7 work packages
• When you need agent-sized context windows for implementation handoffs
• When deciding build order and identifying blocking dependencies
• When breaking large features into executable EPICs
• When preparing EPICs to feed Test Planning and TEST- entries

Best practices

• Keep each EPIC as a cognitive boundary: describe the goal in one sentence
• Size targets: aim for 3–5 APIs, 2–4 DBT tables, 1–2 user journeys per EPIC
• Monitor context during sessions; pause and checkpoint if loaded context approaches limits
• Give every EPIC ID references (BR-, API-, DBT-, ARC-, FEA-) and a single branch owner
• Sequence infrastructure and core data model EPICs before feature work
• Use the 5-phase structure (Plan, Design, Build, Validate, Finish) and include explicit resume instructions

Example use cases

• Create EPICs for auth, core user model, and onboarding as separate context-limited units
• Split a monolithic feature into domain-aligned EPICs when >10 APIs or >5 DBT tables are required
• Generate branch-named EPIC entries with Phase A checkpoints to start work immediately
• Produce EPIC context capsules so an AI agent can complete a session without external lookups
• Output EPICs with pre-load checklists to drive precise Test Planning (TEST- entries)