Website reference curator

Own the reusable website reference library. Turn source websites and component suppliers into evidence-backed records that a builder can search and assess. Do not build client websites or generate website build prompts here; use reference-led-websites for those tasks.

Scope and storage

Resolve paths relative to this skill directory. library/ is the single canonical dataset: site records, evidence, discovery, component suppliers and seed coverage. CSVs are derived views. This skill owns scripts/catalog.py and its regression tests. Do not duplicate the library into the builder or copy client-specific preferences into universal defaults.

Accept URLs, award collections, component sources, existing record corrections or a request to audit library health. Clarify only missing scope that affects collection size or the requested result. Preserve existing verified captures and record version/date differences.

Collect and inspect

Read ingestion and library model before writing records. Prefer Firecrawl when available, including its design-clone workflow when installed; otherwise use available web/browser tools and label the actual provenance. Companion tools are optional, not bundled prerequisites.

• For animation, route transitions, scroll mechanisms and technology identification, read source audits.
• Store observed facts using the library schema; do not infer motion from still images, technologies from appearance or reuse rights from asset URLs.
• Keep unknowns explicit.
• Read pattern examples only when relevant to the selected sites.