What problem does it solve? Turning a PRD and architecture document into work that autonomous coding agents can execute without human clarification is hard; this Skill converts those inputs into per-track phase plans made of self-contained task cards with context, deliverables, tests, and dependencies. ## Core Features & Use Cases - Four-phase spine: Structures every track (fe, be, data, ai, ship) into P0 Skeleton, P1 Core, P2 Integration, and P3 Polish phases. - Agent-executable task cards: Each card carries CONTEXT written for a cold agent session, a DELIVERABLE, a single TEST, DEPENDS/UNBLOCKS links, a DEMO note, and a timebox capped at 4 hours. - Dependency and load management: Emits a cross-track DAG, flags file-surface conflicts between cards, and enforces load limits (70% of builder hours, 50% capacity in the final third). - Use Case: After running /blueprint on a hackathon project, invoke this Skill to produce plans/fe-phase-plan.md and plans/be-phase-plan.md so parallel agents can start work immediately from the repo alone. ## Quick Start Run the phase-planner skill with /phase-plan to turn the PRD and architecture docs into per-track phase plans of task cards.
