What problem does it solve? Starting a development project from scattered planning documents (PRDs, SRS, WBS, RFCs, ADRs) often leads to missed requirements, silent cross-document conflicts, and lost context when sessions end. This Skill turns any set of specification documents into a validated, executable project plan and then builds it task by task with persistent memory. ## Core Features & Use Cases - Document-agnostic intake and validation: Classifies any planning documents into six roles (PRODUCT_INTENT, BEHAVIOR_SPEC, ARCHITECTURE, UX_SPEC, WORK_PLAN, OPS_CONSTRAINTS), cross-checks them on at least four consistency axes, and surfaces BLOCKERs before any code is written. - Project Decision Register and Context Pack: Extracts every binding decision with a traceable source, then generates CLAUDE.md, AGENTS.md, .cursorrules, or copilot-instructions.md plus a project-fit agent roster so any AI tool can pick up the project cold. - Seven-stage execution loop with memory: Runs each task through PLAN, TEST, IMPLEMENT, REVIEW, VERIFY, REMEMBER, IMPROVE, auto-invoking specialist sibling skills, and maintains /project-memory/ files plus Session Snapshots for lossless resume. - Use Case: You have a PRD, an SRS, and a WBS for a new SaaS app. Hand them over, get a validated decision register and context pack, then watch the project get built task by task with test-first discipline and resumable snapshots. ## Quick Start Ask the assistant to kick off a new project from your uploaded PRD, SRS, and WBS documents and build the execution context pack before writing code.
