Layer 2: Skill Metadata — The Cheap, Always-Scanned Descriptor
~11 min read
Layer 2 is just the YAML frontmatter of every skill — 2-3 lines, under 200 tokens — cheap enough that Claude can scan ALL of a project's skills every turn to decide relevance.
Layer 2: Skill Metadata — The Cheap, Always-Scanned Descriptor is a Pro topic
Sign in, then upgrade to Pro or Power to unlock this topic and the full AI Engineering curriculum.
Key points
- •Layer 2, Skill Metadata, is just the YAML frontmatter — about 2-3 lines, under 200 tokens per skill
- •It's deliberately tiny so Claude can scan EVERY available skill's metadata to check relevance, without needing to load any full skill body first
- •This tiny-and-always-scanned design is exactly what lets the system scale to hundreds of skills without blowing the context budget on triage alone
- •Layer 2's job is triage: decide which skill(s), out of potentially hundreds, are actually relevant to the current request
- •Only skills selected by this metadata scan go on to trigger Layer 3 (loading their full instructions) — the expensive step only happens for genuinely relevant skills