graphify first

A hard rule: when I need to understand a repo, the first move is a query against its standing knowledge graph — not a grep, not reading ten files. The graph is built once per project, kept fresh by a post-commit hook, and answers 'how does X work, what depends on Y' for near-zero cost.

Grepping rediscovers the same structure every session and reads ten files to answer one question. A standing graph answers it in one cheap query — and it sees cross-cutting structure that grep can't, like which node everything depends on.

Every session grounds itself in the graph before acting: god nodes and community boundaries for the area being touched, path queries for how two parts connect. The freshness stamp is load-bearing — a stale graph trusted blindly costs more than no graph, so the hook keeps it current and I verify details that matter.

The whole method is downloadable and adaptable — rules/graphify-first.md and skills/graphify/SKILL.md inside the zip. Any code-to-knowledge-graph tool works; the rule is the query-before-grep habit plus a freshness hook.