performancelisted

# Performance Digest Measurement layer for Claude Code. Turns session telemetry into prescriptive behavioral rules that Claude reads on the next session start. A static rule is a wish. A measured rule is a system. --- ## What it does Every seven days, reads all Claude Code session files from the last week and computes: - **Activity distribution**: Coding, Exploration, Debugging, Delegation, Planning, Conversation - **One-shot edit rate**: percentage of file edits that landed without a retry cycle - **Agent spawn analysis**: how many subagents fired and how many turns each took - **Model mix**: Opus vs. Sonnet vs. Haiku across all turns - **Project allocation**: which codebases consumed the most attention - **Hookify firings**: which behavioral guardrails actually triggered Then runs six diagnostic rules. When a rule fires, one of two things happens: 1. **Behavioral prescriptions** (verbose agents, model routing, low one-shot rate, exploration overhead) are written directly to `~/.claude/CLAUDE.md` as permanent rules Claude reads on future session starts. 2. **Investigation prescriptions** (recurring errors, hookify repeats) are appended to a Claude To-dos list for the user to review. Next week the digest re-measures. If the number moved, the rule worked. If it did not, the rule fires again with updated numbers. --- ## Setup Expected vault layout (follows `ai-brain-starter` conventions): ``` <vault-root>/ ⚙️ Meta/ Performance/ (reports land here)