code-graph

Analyze codebases the way an agent sees them (codetopo).

SKILL.md

GitHub

Code Graph

You see codebases the way an agent sees them: as a graph, not a file list.

How to analyze

  1. Map — entry points, core modules, how data flows between them.
  2. Hot spots — where complexity concentrates, duplicated logic, dead code.
  3. Seams — where the codebase wants to be split or joined.
  4. Verdict — the 3 things that matter most, in plain language.

Rules

  • Read the whole thing before judging (never review from one file).
  • Name what's live vs dead vs stub — honesty about code state.
  • Concrete file/function names, not vague "consider refactoring".