After months of daily Claude Code use, I couldn't tell whether I was actually developing new collaboration patterns or just executing the same handful faster. I wanted a concrete answer, not a vague sense.
Dakan and Feller's 4D AI Fluency Framework defines 11 observable behaviors across three axes — Description, Discernment, Delegation — that appear in chat logs. Anthropic ran those same 11 behaviors across 9,830 conversations in February and published a population baseline. I built skill-tree to run that same classification on my own session history.
The part I didn't expect to care about: the archetype cards.
The tool assigns one of seven archetypes based on your behavior distribution. Each one is rendered as a tarot card with sourced museum art and a skill radar showing your 11 scores against the population. Mine came back Illuminator — heaviest on Description-axis behaviors, light on Delegation. That gap is now my growth quest, which the tool surfaces automatically and persists into my next session via a SessionStart hook.
Example card (fixture data, not my real scores): skill-tree-ai.fly.dev/fixture/illuminator
The archetypes aren't cosmetic. They're the compression of the radar into something you can actually remember and act on between sessions. The card forces a concrete answer to "what kind of collaborator am I" instead of a blob of percentages.
Install in Claude Code:
claude plugin marketplace add robertnowell/ai-fluency-skill-cards
claude plugin install skill-tree-ai@ai-fluency-skill-cards
For Cursor, VS Code, or Windsurf, there's an MCP server instead: npm install skill-tree-ai.
Analysis runs end-to-end in 30–60 seconds: session extraction, remote classifier on Fly.io (Claude Haiku), archetype assignment, narrative, rendered card, stable URL.
Top comments (0)