Check reference / Token & Context Cost
TOKEN-01 Body under line/token budget
Deterministic — runs on every scan, no LLM or network needed.
Why it matters
The full SKILL.md is injected into context every time the skill fires, so every line is a recurring token (and dollar) cost on every invocation — and a bloated file risks your actual instructions being diluted or truncated. The official guidance is ~500 lines; this check warns above 500 and fails above 750.
How to fix it
Cut SKILL.md to under 500 lines by moving heavy reference material into supporting files and linking to them (progressive disclosure), so the model loads that material only when it needs it. The evidence shows your estimated tokens and per-load cost at a $3/Mtok reference rate.