Anthropologists call it competence plateau: you get fast, you stop noticing what you're not doing. After months of daily Claude Code sessions, I couldn't answer a simple question — of the 11 observable collaboration behaviors Anthropic measured across 9,830 real Claude conversations in their February 2026 AI Fluency Index, which ones do I actually use, and which have I never touched?
So I built skill-tree to answer that on my own session history.
The tool installs as a Claude Code plugin:
claude plugin marketplace add robertnowell/ai-fluency-skill-cards && claude plugin install skill-tree-ai@ai-fluency-skill-cards
Once installed, it runs a 7-step orchestration — finds your session files, extracts user messages, sends them to a remote classifier (Claude Haiku on Fly.io), maps your behavior distribution against the population baseline from Dakan & Feller's 4D AI Fluency Framework, assigns you one of seven archetype cards, then returns a stable URL to your rendered results. End-to-end takes 30–60 seconds.
The archetype cards are rendered as tarot cards using curated museum art. You can see a live example at skill-tree-ai.fly.dev/fixture/illuminator. The card isn't the point — the behavior breakdown is. Specifically: which of the 11 behaviors you've never triggered across your recent sessions becomes your growth quest for the next one, persisted via a SessionStart hook so Claude reminds you at the top of each new session.
The part that surprised me most when I ran it on my own history: I was heavily concentrated in two of the three measurable axes (Description and Delegation), and had basically zero signal on Discernment behaviors. The framework distinguishes these from a fourth axis, Diligence, which isn't observable from chat logs alone.
Also works in Cowork via skill-tree-ai.zip, and as an MCP server (npm package skill-tree-ai) for Cursor, VS Code, and Windsurf.
Top comments (0)