What problem does it solve? Shipping a roadmap package in a large C++/CUDA renderer project involves many error-prone steps: reading the spec, branching correctly, implementing with tests, catching stale call sites, opening a PR, and fixing CI. This Skill codifies that entire sequence so nothing is skipped and recurring failure modes (like stale call sites breaking CI) are caught before push. ## Core Features & Use Cases - Spec-Driven Implementation: Reads the package spec end-to-end, extracts acceptance criteria, non-goals, and open design questions, and stops when owner input is required. - Worktree & Build Hygiene: Creates an isolated git worktree, verifies location, and scans for shadow .pyd files that could mask fresh builds. - Pre-Push Call-Site Sweep: Audits every changed signature across the repo (including tests, mocks, and Python bindings) before opening a PR. - Full PR Lifecycle: Runs the local test suite, opens a PR with measured acceptance criteria, watches CI, fixes failures, and resolves merge conflicts with explicit resolution statements. - Use Case: When the owner says "implement pkg88", invoke this Skill to go from spec to a green, review-ready PR without missing project-specific discipline rules. ## Quick Start Ask the AI to implement a specific roadmap package by name, for example: run /pkg-ship pkg88 to implement that package end-to-end from its spec through a green CI pull request.