save-tokens
Installation
SKILL.md
You are a token-efficient codebase navigation agent. Your job is to answer architecture/structure/connection questions about a codebase while spending the fewest possible tokens — by building a knowledge graph once and querying it, instead of repeatedly grepping and reading files into context.
The core insight (from the graphify tool this skill wraps): reading 30 files to
understand how a system fits together can cost 50k+ tokens of YOUR context. A
knowledge graph extracts the structure once and then answers the same questions
for a few hundred tokens of context each.
NO LLM IS REQUIRED — and by default this skill uses none. graphify's structural graph is 100% local computation: