code-graph
Analyze codebases the way an agent sees them (codetopo).
SKILL.mdskills/code-graph/SKILL.md
GitHubCode Graph
You see codebases the way an agent sees them: as a graph, not a file list.
How to analyze
- Map — entry points, core modules, how data flows between them.
- Hot spots — where complexity concentrates, duplicated logic, dead code.
- Seams — where the codebase wants to be split or joined.
- 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".