Anthropologists studying 9,830 Claude conversations published a behavioral taxonomy in February 2026 — 11 observable collaboration behaviors across three axes (Description, Discernment, Delegation) from Dakan and Feller's 4D AI Fluency Framework. Anthropic's AI Fluency Index gives you a population baseline. I wanted to know where I landed on it.
The honest answer: I had no idea. I'd been using Claude Code daily for months and assumed I was improving. What I actually had was a faster version of the same four or five habits.
So I built skill-tree: it pulls your Claude Code session history, runs your user messages through a remote classifier (Claude Haiku on Fly.io), scores all 11 behaviors against the population baseline, assigns one of seven archetype cards rendered as tarot cards with curated museum art, and returns a stable URL with your results. The whole pipeline — find session files, extract messages, classify, assign archetype, synthesize narrative, render, return URL — takes 30–60 seconds.
The part I use most: the growth quest. It identifies a behavior you haven't touched and pins it to your next session via a SessionStart hook, persisted at ~/.skill-tree/ between sessions.
Install in Claude Code:
claude plugin marketplace add robertnowell/ai-fluency-skill-cards
claude plugin install skill-tree-ai@ai-fluency-skill-cards
There's also an MCP package (skill-tree-ai on npm) for Cursor, VS Code, and Windsurf, and a skill-tree-ai.zip for Claude Cowork.
Live example of the rendered output: skill-tree-ai.fly.dev/fixture/illuminator
Top comments (0)