After months of daily Claude Code sessions, I noticed my prompts had converged. Same patterns: describe the bug, paste the error, ask for a fix. Fast, functional, flat. No delegation, no discernment — just description on repeat.
Dakan & Feller's 4D AI Fluency Framework breaks AI collaboration into 11 observable behaviors across three axes you can see in chat logs: Description, Discernment, and Delegation. Anthropic ran those behaviors against 9,830 Claude conversations in February 2026 and published the distribution. I wanted to know where I landed — and more importantly, which behaviors I had never tried.
So I built skill-tree. It reads your Claude Code or Cowork session history, classifies all 11 behaviors, assigns one of seven archetype cards (rendered as tarot cards with museum art), and picks a growth quest for your next session.
The quest selection is the part I thought hardest about. The naive approach is to surface your weakest behavior from the ones you've already tried. That just optimizes within your existing repertoire. skill-tree targets behaviors you haven't touched at all — behaviors with zero representation in your session history. Local optima are comfortable; that's why you're in one.
A concrete example: if your sessions show strong Description and some Discernment but zero Delegation, the quest pushes you to try something like asking Claude to decompose a problem autonomously before you've framed it. That's a different cognitive mode than refining a prompt you've already drafted. The discomfort is the signal.
The quest persists across sessions via a SessionStart hook, stored at ~/.skill-tree/ in Claude Code. You see it at the start of each session until you've demonstrated the target behavior, then it updates.
The full analysis — classifier, archetype card, narrative — runs in 30–60 seconds. A live example card is at skill-tree-ai.fly.dev/fixture/illuminator.
Install in Claude Code:
claude plugin marketplace add robertnowell/ai-fluency-skill-cards
claude plugin install skill-tree-ai@ai-fluency-skill-cards
Also available as an MCP server (npm install skill-tree-ai) for Cursor, VS Code, and Windsurf.
Top comments (0)