Anthropologic's February 2026 study classified 11 observable collaboration behaviors across 9,830 Claude conversations — things like delegation depth, discernment quality, how you frame ambiguous tasks. I'd been using Claude Code every day for months, but when I looked at that taxonomy honestly, I could count the behaviors I actually varied on one hand.
skill-tree runs against your local Claude Code or Cowork session history, classifies the same 11 behaviors using the same framework, and maps where you cluster.
The classifier is Claude Haiku on Fly.io, takes 30–60 seconds end-to-end across a 7-step pipeline: find session files, extract user messages, classify, assign archetype, synthesize narrative, render, return a stable URL. The archetypes are rendered as tarot cards with museum art — seven total, one assigned per analysis. Live example at skill-tree-ai.fly.dev/fixture/illuminator.
The part I actually built it for: it picks one behavior you haven't touched and sets it as a growth quest that persists into your next session via the SessionStart hook. Not a generic tip — a specific behavior from the 4D AI Fluency Framework (Description, Discernment, Delegation axes) that your own log shows you've skipped.
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: skill-tree-ai) for Cursor, VS Code, Windsurf.
github.com/robertnowell/skill-tree
Top comments (0)