claude-extension-logging

To capture Claude extension traffic for debugging, you instrument the extension and shim by appending JSON lines to specific files before dispatching messages, ensuring all directional traffic logs are consistently recorded.

To log shim vs extension message paths separately, apply distinct naming conventions using `claude_*` for extension traffic and `claude_shim_*` for shim traffic, allowing you to differentiate and compare communication streams.

CLI response wrapping for Claude extension logging mirrors the extension envelope structure, ensuring command-line interface responses are formatted consistently with extension messages for unified observability and comparison.

Yes, you can use this logging approach to compare Claude traffic with other observability systems like Codex, as it captures all message exchanges in structured JSON lines to facilitate cross-system comparison.

Claude extension logging requires no external dependencies, operating by appending JSON lines to specific files and wrapping CLI responses without needing additional packages or libraries.

You need directional traffic logs for the Claude ecosystem when debugging issues related to Claude's interaction with its UI or command-line interface, ensuring no message data is lost between the extension, shim, and webview.