Reads JSONL logs from Claude Code and parses conversations.json from Claude.ai exports.
Filters to human-authored prompts. Excludes system commands and tool outputs.
Regex pattern matching to count politeness markers, questions, and commands.
Maps Engagement and Politeness scores onto a 2×2 matrix.
Prompts — your messages · Conversations — chat sessions · Words — total typed
Avg Turns — per convo · Quick Asks — 1-3 turns · Deep Dives — 11+ turns
Peak Hour/Day — most active · Night Owl — 11pm-4am %
Politeness — please/thanks · Backtrack — actually/wait · Questions — ending ? · Commands — action verbs
Engagement = (Question + Backtrack) ÷ 2Politeness = Index − (Command × 0.5)
Based on where you fall on the Engagement (x-axis) and Politeness (y-axis) scales:
Thresholds: Engagement ≥ 12, Politeness Score ≥ 4.5