After months of daily Claude Code use, I noticed my sessions had settled into a rut: describe a problem, get code, tweak it, repeat. Fast, but narrow. I wanted to know which collaboration behaviors I was skipping entirely — not just vibes, but against something measurable.
Anthropic published a study in February (9,830 Claude conversations) that classifies 11 observable behaviors across three axes — Description, Discernment, Delegation. I built skill-tree to run that same classification on your own session history.
The output is a tarot-style archetype card, rendered with curated museum art, plus a skill radar showing which of the 11 behaviors appear in your sessions and which you haven't touched. Seven archetypes total: Illuminator, Navigator, Alchemist, and four others. Each maps to a specific behavioral cluster from the Dakan & Feller 4D AI Fluency Framework. Live example of what the card looks like: skill-tree-ai.fly.dev/fixture/illuminator
The part I find most useful isn't the card — it's the growth quest. The tool picks one behavior absent from your recent sessions and pins it as a target for your next one. It persists via a ~/.skill-tree/ file and surfaces again at session start through a SessionStart hook.
Installs directly in Claude Code:
claude plugin marketplace add robertnowell/ai-fluency-skill-cards
claude plugin install skill-tree-ai@ai-fluency-skill-cards
Also works in Cowork (via skill-tree-ai.zip) and in Cursor, VS Code, or Windsurf via the skill-tree-ai MCP npm package. Full analysis — session extraction, remote classifier on Fly.io, archetype assignment, card render — runs in about 30–60 seconds and returns a stable URL.
Top comments (0)