[

[
"# The behavior you never try is the one the growth quest targets\n\nAnthropic’s AI Fluency Index (February 2026, 9,830 conversations) classifies 11 observable collaboration behaviors â€" things like explicit goal-setting, iterative refinement, and constraint specification. Most people who use Claude Code daily are strong on two or three of these and have never touched the rest. That's a local optimum, not fluency.\n\nskill-tree reads your Claude Code session history, classifies which of those 11 behaviors actually appear in your messages, and then picks a growth quest from the behaviors with the lowest or zero representation in your data. Not the behaviors you're weak at â€" the ones you don't reach for at all. The anti-local-optimum rule.\n\nThe classification runs on Claude Haiku on Fly.io. The full 7-step pipeline â€" find session files, extract user messages, classify, assign archetype, synthesize narrative, render, return URL â€" takes 30â€"60 seconds. What comes back is a stable URL with a tarot-style archetype card (museum art, seven possible archetypes) and one specific behavior flagged as the next session's quest. A live example is at skill-tree-ai.fly.dev/fixture/illuminator.\n\nThe quest persists via SessionStart hook, so it surfaces at the top of your next Claude Code session without you having to remember to check.\n\nInstall:\n\n

bash\nclaude plugin marketplace add robertnowell/ai-fluency-skill-cards\nclaude plugin install skill-tree-ai@ai-fluency-skill-cards\n

\n\nOr as an MCP server for Cursor, VS Code, or Windsurf:\n\n

bash\nnpm install skill-tree-ai\n

\n\nThe behavior taxonomy and three-axis model (Description, Discernment, Delegation â€" Diligence is the fourth axis but isn't recoverable from chat logs) come from Dakan & Feller's 4D AI Fluency Framework. The population baseline is Anthropic’s February 2026 study. Your numbers are measured against that, not against some abstract ideal.\n\nhttps://github.com/robertnowell/ai-fluency-skill-cards",