What problem does it solve? Turning an approved ICEA and Tech Spec into actual code is error-prone: developers drift from the spec, skip acceptance criteria, or write code that fails review. This Skill implements an approved ICEA end-to-end — generating code per layer, gating it through a critic and an AC-completeness goal-loop, and only writing to disk after explicit approval. ## Core Features & Use Cases - Spec-driven code generation: Reads the approved ICEA, Tech Spec, and tracker for an Azure DevOps work item, then generates persistence, service, API, UI, and test code in dependency order, matching the project's actual stack (.NET, Java/Spring, Python, Node.js, Angular/React) and per-project .NET version. - Pre-write quality gates: Runs an auto-critic for ICEA/Tech-Spec traceability and a bounded goal-loop that scores completion against every acceptance criterion and planned file change before anything touches disk. - Write Gate and audit trail: Presents all pending files for explicit APPROVE, blocks boundary-crossing writes outside the repo, updates trackers, and appends tamper-evident audit rows for every revision, fix, and completion event. - Use Case: A developer runs IMPLEMENT ADO-1847 after the ICEA is approved; the Skill generates the Dapper repository, service, controller, and xUnit tests, passes the critic and goal-loop, and writes everything only after the developer approves the file list. ## Quick Start Ask the AI to implement an approved work item by saying: implement ADO-1847 and generate the code, tests, and tracker updates for my approved ICEA.
