transcript-extract

Extract structured events from Claude Code session transcripts as JSONL.

Updated Feb 23, 2026
One-click install
npx skills add https://github.com/ASRagab/asragab-claude-marketplace --skill transcript-extract
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: transcript-extract
Source: https://github.com/ASRagab/asragab-claude-marketplace/tree/main/plugins/skill-eval/skills/transcript-extract
Command: npx skills add https://github.com/ASRagab/asragab-claude-marketplace --skill transcript-extract

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires bun, and includes scripts (resource) components.

What problem does it solve?

Extracts structured, machine-readable events from Claude Code session transcripts stored in ~/.claude/projects/ for easier analytics and auditing.

Core Features & Use Cases

  • Structured event extraction that outputs JSONL lines representing user messages, tool uses, tool results, thinking blocks, progress events, and subagent transcripts.
  • Flexible filtering by session or project, optional inclusion of subagents, and multiple output formats (jsonl or summary) for downstream analytics.

Quick Start

Run bun scripts/transcript-extract.ts --since 7d -o events.jsonl to extract recent transcript events.

Frequently Asked Questions about transcript-extract

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I extract structured events from Claude Code session transcripts?

Extract structured events from Claude Code session transcripts by running a Bun script that reads directly from ~/.claude/projects/ and processes user messages, tool uses, and thinking blocks into machine-readable JSONL lines for downstream analytics.

What types of transcript events can I parse from Claude Code project logs?

Parsing Claude Code project logs extracts structured events including user messages, tool uses, tool results, thinking blocks, progress events, and subagent transcripts. You can optionally include subagent transcripts when filtering sessions.

Do I need Bun runtime to process Claude Code transcripts into JSONL format?

Yes, you need the Bun runtime installed to execute the transcript extraction script. The Skill uses Bun to process transcripts directly from ~/.claude/projects/ without requiring any external dependencies.

Can I filter transcript extraction by specific project or session ID?

Yes, you can filter transcript extraction by specific project or session. The script processes transcripts across sessions and projects with flexible filtering options, allowing you to target recent activity using time-based flags like --since 7d.

What output formats are available when extracting Claude Code session data?

Extracting Claude Code session data supports JSONL and summary output formats. You specify the desired format and destination file using command-line flags when running the extraction script for downstream analytics.

What's the best way to audit Claude Code tool usage across multiple projects?

The best way to audit Claude Code tool usage across projects is extracting structured events from session transcripts in ~/.claude/projects/. The script processes multiple sessions, capturing tool uses and results into a consolidated JSONL file for analysis.