STOP. READ THIS ENTIRE SKILL.MD BEFORE CALLING ANY ENDPOINT.

Conversation Lab

Self-improving convergence loop for SPARTA stress test conversations. Follows the Shadow-Lego pattern: run -> assess -> diagnose -> fix -> re-run.

Commands

Command	Description
`diagnose [file]`	Structured JSON diagnosis for `/assess` (use `--format md` for Rich tables)
`report [file]`	Full markdown report with conversation transcripts (like `/tmp/conversation_flows_*.md`)
`data [file]`	Export JSON chart data for `/create-figure` (radar, heatmap, grade distribution)
`converge [file]`	Re-run unsatisfied sessions until personas are happy or ceiling hit
`optimize`	Analyze session data + episodic archives to recommend optimal turn counts
`status`	Show convergence state (running/complete/stalled)