What problem does it solve?
Service teams maintaining Azure SDK for Java packages need a repeatable way to encode package-specific tribal knowledge (customization patterns, generated vs hand-written code boundaries, regeneration pitfalls) into discoverable agent skills, but writing a well-structured SKILL.md from scratch is error-prone and inconsistent.
Core Features & Use Cases
- Package Scanning: Detects TypeSpec generation, customization files, generated vs hand-written sources, service versions, and test setup to build a package profile.
- Guided Scaffolding: Generates a SKILL.md with required sections (Common Pitfalls, Architecture, After Regeneration, Testing) plus architecture.md and customizations.md reference files.
- Structural Validation: Runs vally lint checks (spec-compliance, valid-refs, orphan-files) and token budget verification before finalizing.
- Use Case: A Cosmos SDK engineer asks the agent to bootstrap a skill for azure-cosmos; the wizard scans the package, drafts the skill with package-specific pitfalls, validates it, and places it under the package's .github/skills/ directory for automatic discovery.
Quick Start
Ask the agent to create a package skill for your Azure SDK package, for example by saying "create a package skill for sdk/search/azure-search-documents".