Layer 3: Active Skill Context — Full Instructions, Loaded On Demand
~13 min read
Layer 3 loads a skill's full SKILL.md body only once it's been selected as relevant — and supporting files (scripts, templates) go further still, never entering context at all until actually invoked.
Layer 3: Active Skill Context — Full Instructions, Loaded On Demand 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 3, Active Skill Context, loads a skill's full SKILL.md body and associated documentation only once Layer 2 has flagged it as relevant
- •The YAML frontmatter (Layer 2) and Skill Body (Layer 3) are two sections of the SAME skill.md file — not separate files, just loaded at different times
- •Supporting files (scripts, templates, reference docs) go further: they're never loaded into context at all, only fetched directly when actually needed, costing zero tokens
- •Each layer's cost scales differently: Layer 1 with the project (roughly constant), Layer 2 with total skill count (but tiny each), Layer 3 with only currently-active skills
- •This is exactly what lets the whole architecture support hundreds of skills without breaching context limits — cost is proportional to relevance and use, not library size