Asset Library Main Entry

What problem it solves

When a user says “turn my materials into an asset library, query it, and export it” without naming the current stage—raw materials ingest, build-asset-library, maintain-asset-library, query-assets, check-lifecycle, or build-package—running one subflow directly can skip prerequisites. This entry skill separates intent, available facts, and stage boundaries before execution, preventing multiple subskills from being chained into one run.

How it works

• Stage detection: It first checks whether the user is describing the whole pipeline or a specific stage; if a stage is already explicit, it does not force the whole workflow.
• Single routing step: It selects the minimal default MVP subskill among ingest-materials, build-asset-library, maintain-asset-library, query-assets, check-lifecycle, and build-package; submit-demo is treated as an optional advanced extension, not a default route.
• Input and boundary check: It captures current intent and available facts, such as local file paths, library_id, goal, package_plan_id, or package_id, then reports missing inputs, stage boundaries, and a short next_recommended_skill name.
• Readiness first: For first install, missing keys, missing profile, missing MCP/skills registration, or unclear semantic-search/old-library indexing status, it defers to references/install.md instead of assuming runtime readiness.

Scope and limits

It does not directly call MCP tools, OCR, SQLite, docgen, or browser actions, and it does not rewrite subskill input/output contracts. It is suitable for routing an asset-library pipeline, not for performing ingest, library construction, query, or packaging itself.