{
  "name": "claude-performance",
  "description": "Weekly performance digest for Claude Code. Reads JSONL session data, computes six effectiveness metrics (activity mix, one-shot edit rate, subagent turn count, model mix, project allocation, hookify firings), fires diagnostic rules, and writes behavioral rules into CLAUDE.md so future sessions adapt. Measurement layer for self-improving AI workflows.",
  "version": "1.0.0",
  "author": {
    "name": "Mycelium AI",
    "url": "https://myceliumai.co"
  },
  "repository": "https://github.com/adelaidasofia/claude-performance",
  "license": "MIT"
}