I've been using Claude Code daily for months. At some point I noticed I was fast — but I couldn't tell if fast meant skilled or just habitual. Same prompting patterns, same delegation moves, same blind spots I never examined.
Anthropic published a study in February 2026 covering 9,830 Claude conversations, measuring 11 observable collaboration behaviors across their user base. I wanted to know what my distribution looked like against that baseline, and specifically which behaviors I was skipping entirely.
skill-tree analyzes your Claude Code or Cowork session history, classifies those same 11 behaviors using the 4D AI Fluency Framework (Dakan & Feller), assigns one of seven archetype cards, and picks a behavior you haven't touched as a growth quest for your next session.
The behavior taxonomy runs on three visible axes — Description, Discernment, Delegation — plus a fourth (Diligence) that doesn't surface in chat logs. The classifier runs on Fly.io using Claude Haiku, takes 30–60 seconds end-to-end across seven steps: find session files, extract user messages, classify, assign archetype, synthesize narrative, render, return a stable URL. The cards are rendered as tarot cards with curated museum art. You can see a live example at skill-tree-ai.fly.dev/fixture/illuminator.
The design detail worth explaining: Cowork's $HOME is ephemeral — anything written there doesn't survive across sessions. Growth quests need to persist, so the tool uses ~/.skill-tree/ on Claude Code and $CLAUDE_PLUGIN_ROOT/.user-state/ on Cowork. Same tool, two state paths, different reasons.
Install in Claude Code:
claude plugin marketplace add robertnowell/ai-fluency-skill-cards
claude plugin install skill-tree-ai@ai-fluency-skill-cards
Cowork users: skill-tree-ai.zip. Also available as an MCP server (npm: skill-tree-ai) for Cursor, VS Code, and Windsurf.
Top comments (0)