What is considered 'noise' in a Claude Code JSONL file?

Noise includes internal progress hooks, file-history snapshots for undo/redo functionality, and queue operations that have no narrative value to the user.

Is it safe to use for public documentation?

The skill includes a scanner to flag potential PII and secrets (like API keys), but users should always perform a manual review of the flagged content before publishing.

Can this skill extract images from my coding sessions?

Yes, it identifies base64-encoded screenshots embedded in Playwright tool results and extracts them to a local assets directory for external reference.

Where does Claude Code store its conversation history?

Claude Code stores conversations as JSONL files within ~/.claude/projects/. The directory names are encoded versions of your project's absolute path.

How does this skill handle subagent conversations?

It uses a linking procedure to match tool_use IDs from the main conversation with agentId strings found in tool_results, allowing it to locate the corresponding subagent JSONL files.

Claude Code Conversation Parser

Name: Claude Code Conversation Parser
Author: oborchers

byoborchers

•

Analíticas y Monitorización

Extracts high-signal narrative data and clean transcripts from Claude Code's internal JSONL conversation storage.

The Conversation Format skill provides the deterministic logic and regex patterns required to parse Claude Code's complex JSONL storage format. It enables developers to transform raw project logs into structured narratives by filtering out internal system noise (like file snapshots and progress hooks) while isolating critical human-AI interactions. This skill is essential for anyone building documentation, blog posts, or research tools on top of Claude Code, as it handles session boundary detection, subagent transcript linking, and the extraction of embedded visual assets like Playwright screenshots.

Características Principales

01Session boundary detection using UUID shifts and continuation markers

02Recursive subagent linking and transcript retrieval

03Integrated PII and secret scanning for secure content triage

04Automated noise filtering for system metadata and internal snapshots

059 GitHub stars

06Deterministic signal extraction from Claude Code JSONL transcripts

Casos de Uso

01Converting CLI conversation histories into readable technical blog posts

02Auditing AI development workflows for sensitive data or hardcoded secrets

03Extracting Playwright screenshots and tool outputs for visual documentation

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add oborchers/fractional-cto conversation-format

For use in Claude.ai and ChatGPT

Características Principales

01Session boundary detection using UUID shifts and continuation markers

02Recursive subagent linking and transcript retrieval

03Integrated PII and secret scanning for secure content triage

04Automated noise filtering for system metadata and internal snapshots

059 GitHub stars

06Deterministic signal extraction from Claude Code JSONL transcripts

Casos de Uso

01Converting CLI conversation histories into readable technical blog posts

02Auditing AI development workflows for sensitive data or hardcoded secrets

03Extracting Playwright screenshots and tool outputs for visual documentation

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add oborchers/fractional-cto conversation-format

For use in Claude.ai and ChatGPT