Where are the Claude Code session transcripts stored?

Transcripts are stored in JSONL format under the ~/.claude/projects/ directory, using a dash-separated version of your project's file path.

How does this skill help reduce context bloat?

It provides specific tools, like 'transcript-tool bloat', to identify the largest messages and tool results that are consuming unnecessary context space.

Can I run custom queries on my session logs?

Yes, the skill includes a CLI tool that supports a custom 'extract' command, allowing you to run raw jq queries against your transcript files for advanced analysis.

What is the primary purpose of the Transcript Analyzer?

It allows developers to inspect and analyze Claude Code session transcripts to debug plugins and understand exactly how context and tokens are being utilized during a session.

Does it support tracking sub-agents?

Yes, the analyzer can distinguish between main conversations and sub-agents by filtering the 'isSidechain' attribute within the transcript records.

Transcript Analyzer

Name: Transcript Analyzer
Author: bfreis

bybfreis

•

分析与监控

Analyzes Claude Code session transcripts to debug plugin behavior, optimize context usage, and trace tool execution flows.

关于

The Transcript Analyzer is a specialized skill for Claude Code that provides deep visibility into LLM interaction logs stored in JSONL format. It allows developers to parse session metadata, track file history snapshots, and audit token usage across user and assistant messages. By identifying sources of context bloat and investigating unexpected tool behavior through a dedicated CLI tool, this skill helps developers optimize their AI-driven workflows, minimize API costs, and ensure custom plugins are functioning as intended.

主要功能

Debug plugin logic by tracing specific tool invocations and results
Parse JSONL session transcripts for detailed execution tracing
Analyze message structure including assistant thinking and tool calls
Support for custom jq queries to perform advanced log data extraction
Identify and mitigate sources of context bloat and excessive token usage
1 GitHub stars

使用场景

Debugging failing or unexpected behavior in custom Claude Code plugins
Auditing the sequence of tool calls to verify complex project-wide execution flows
Monitoring token consumption to reduce API costs and improve response speed

关于

主要功能

Debug plugin logic by tracing specific tool invocations and results
Parse JSONL session transcripts for detailed execution tracing
Analyze message structure including assistant thinking and tool calls
Support for custom jq queries to perform advanced log data extraction
Identify and mitigate sources of context bloat and excessive token usage
1 GitHub stars

使用场景

Debugging failing or unexpected behavior in custom Claude Code plugins
Auditing the sequence of tool calls to verify complex project-wide execution flows
Monitoring token consumption to reduce API costs and improve response speed