cyberneticlibrary
← library

Analyze Claude Code sessions

braintrust-analyzeskillsetup L2★0
David-Li0406/meta-skill-evloving ↗
What it does

Analyze Claude Code session patterns via Braintrust tracing

Best for

Retrospective productivity review and loop-detection when debugging why a session got stuck or repeating the same tool calls

Inputs
  • · BRAINTRUST_API_KEY
  • · optional session ID or time range
Outputs
  • · tool usage breakdown
  • · agent spawn log
  • · skill activation sequence
  • · token consumption estimate
  • · loop detection (repeated tool calls)
Requires
  • · Braintrust API
  • · braintrust-claude-plugin
Preconditions

Braintrust tracing enabled in project; BRAINTRUST_API_KEY in .env

Failure modes

Missing API key causes silent failure; tracing disabled yields empty data; session ID mismatch returns nothing

Trust signals
  • · Python script with named CLI commands (--last-session, --detect-loops, --weekly-summary)
  • · Real output example showing session breakdown format
  • · Clear explanation of loop detection as finding same tool >5 times
  • · Integration with Braintrust platform (established observability service)