What problem does it solve? AI-generated specs are often technically correct but product-misaligned because agents lack business context like personas, hard business invariants, and planned scope. This Skill captures that human knowledge in a shared registry stored outside the worktree so every downstream spec starts from the same product truth. ## Core Features & Use Cases - Structured product registry: Generates and interactively fills product_context.md (8 mandatory sections), roadmap.md, and the manual portion of architecture.md under $CHE_PROJECT_DIR. - Three execution modes: --show for read-only summaries, default interactive mode for section-by-section confirmed updates, and --bootstrap to scaffold empty [PENDING] templates on first run. - Storage boundary enforcement: Resolves canonical paths via the che CLI, asserts the registry stays outside the user worktree, and writes atomically with audit entries appended to registry.jsonl. - Use Case: Before running /che-spec on a project's first feature, run this Skill to record personas, non-negotiable business rules, integrations, and out-of-scope features so the generated spec injects that context automatically. ## Quick Start Ask the agent to run che-onboarding in bootstrap mode to create the product context, architecture, and roadmap registry files for this project.
