DEV Community

J Now
J Now

Posted on

Claude Code gave me no mirror. I built one.

Anthropics February 2026 study classified 11 observable collaboration behaviors across 9,830 Claude conversations. I had been using Claude Code daily for months and assumed I was improving. When I actually ran my sessions against that taxonomy, I found I was clustering hard around two behaviors and had never once triggered five others.

That gap is what skill-tree surfaces.

It reads your Claude Code or Cowork session history, runs a remote classifier (Claude Haiku on Fly.io, takes 30-60 seconds end-to-end), and scores your sessions against the same 11 behaviors from the AI Fluency Index. Then it assigns one of seven archetype cards — rendered as tarot cards with museum art — and picks a single behavior you haven't touched as a growth quest for your next session. The quest persists via a SessionStart hook so Claude Code surfaces it at the top of every new conversation until you try it.

The behavior taxonomy comes from Dakan and Feller's 4D AI Fluency Framework, organized across three axes visible in chat logs: Description, Discernment, and Delegation. A fourth axis (Diligence) exists in the framework but isn't observable from message history, so it's excluded.

Install in Claude Code:

claude plugin marketplace add robertnowell/ai-fluency-skill-cards
claude plugin install skill-tree-ai@ai-fluency-skill-cards
Enter fullscreen mode Exit fullscreen mode

For Cowork, drop in skill-tree-ai.zip. For Cursor, VS Code, or Windsurf, the MCP server is on npm as skill-tree-ai.

Live archetype example: https://skill-tree-ai.fly.dev/fixture/illuminator

Repo: https://github.com/robertnowell/skill-tree

Top comments (0)