Anthropologists studying 9,830 Claude conversations published a taxonomy of 11 observable AI collaboration behaviors in February — things like iterative refinement, context-setting, constraint specification. When I read that paper I realized I had no idea which of those 11 I was actually doing in my own sessions versus which ones I'd never touched.
skill-tree classifies your Claude Code or Cowork session history against that same taxonomy and assigns you one of seven archetype cards rendered as tarot cards with museum art. The live example at skill-tree-ai.fly.dev/fixture/illuminator shows what the output looks like.
The part I found most useful isn't the archetype — it's the growth quest. The quest picks the behavior you avoid most, not a behavior you're mediocre at. The reasoning: optimizing a behavior you already use reinforces the same working pattern. A behavior you never touch means an entire collaboration mode you've been leaving on the table.
This is the anti-local-optimum rule baked into the design. If you're fluent at iterative refinement but never use explicit constraint specification, the quest points at constraint specification — not because refinement couldn't be sharper, but because you've already found a stable equilibrium there. The gap is elsewhere.
The quest persists across sessions via a SessionStart hook, so it shows up at the start of the next Claude Code session rather than buried in a report you'll forget to reopen.
claude plugin marketplace add robertnowell/ai-fluency-skill-cards
claude plugin install skill-tree-ai@ai-fluency-skill-cards
The full 7-step orchestration — find session files, extract messages, remote classifier on Fly.io, archetype assignment, narrative synthesis, render, return URL — runs in 30–60 seconds. Also available as an MCP server (npm install skill-tree-ai) for Cursor, VS Code, and Windsurf.
github.com/robertnowell/skill-tree
Top comments (0)