Most feedback loops optimize for your strengths. You get fast at the things you do — then you do those things more because they feel productive. Local optimum, no escape route.
skill-tree does the opposite. After classifying your Claude Code sessions against the 11 behaviors in Anthropic's AI Fluency Index (February 2026, 9,830 conversations studied), it looks at which behaviors you never touch and assigns one of those as a growth quest for your next session.
The taxonomy comes from Dakan & Feller's 4D AI Fluency Framework — Description, Discernment, Delegation, and Diligence. Diligence doesn't appear in chat logs, so the classifier works across the other three axes. If you're heavy on Description (explaining your problem in detail) but never use Discernment behaviors (questioning the output, asking for alternatives, pushing back on reasoning), the quest targets Discernment — not because you're bad at it, but because you've never tried it.
The archetype cards are rendered as tarot cards with curated museum art. You can see a live example at skill-tree-ai.fly.dev/fixture/illuminator. Each run takes 30–60 seconds: session files get extracted, classified remotely via Claude Haiku on Fly.io, assigned to one of seven archetypes, and returned as a stable hosted URL.
The quest persists across sessions via a SessionStart hook, stored at ~/.skill-tree/ in Claude Code so it survives restarts.
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.
github.com/robertnowell/skill-tree
Top comments (0)