When you need a durable knowledge graph of a codebase or mixed corpus that survives beyond the current chat session, this is the routing layer that picks the right execution mode: assistant-native install, local Python build, incremental refresh, query-only follow-up, or structural fallback for markdown-heavy repos. It scopes the corpus before runtime explodes, produces GRAPH_REPORT.md and graph.json artifacts, and knows when to hand off search-only work to codebase-search or wiki work to llm-wiki. The real job is mode classification and honest fallback choice, not dumping every Graphify feature. Works best with Python 3.10+ and the graphifyy PyPI package, supports Claude, Codex, Gemini, and OpenCode assistant installs.
npx -y skills add akillness/oh-my-skills --skill graphify --agent claude-codeInstalls into .claude/skills of the current project.
Select a file.
juliusbrussee/caveman
mattpocock/skills
shadcn/improve
obra/superpowers
forrestchang/andrej-karpathy-skills
vercel-labs/skills