conversation-format

What problem does it solve?

This Skill addresses the challenge of extracting meaningful signal from Claude Code's complex JSONL conversation logs, which are often filled with noise and metadata.

Core Features & Use Cases

• Signal Extraction: Identifies and extracts valuable user and assistant messages, along with crucial metadata like turn duration.
• Noise Filtering: Intelligently discards irrelevant entries such as progress updates and internal system events.
• Content Block Parsing: Differentiates and extracts key information from text, thinking, and tool_use blocks, while selectively handling image data.
• Subagent Linking: Provides a method to link main conversation tool uses to their corresponding subagent transcripts.
• PII Scanning: Includes a mechanism to flag potential Personally Identifiable Information (PII) and secrets within the conversation data.
• Use Case: When analyzing past conversations to understand AI decision-making or to repurpose content, this skill ensures you only process the essential narrative elements, saving significant time and computational resources.

Quick Start

Use the conversation-format skill to parse the conversation transcript located at ~/.claude/projects/example-project/conversation-uuid.jsonl.