I Built a Way to Analyze Claude Code's Logs. It Separated Problems in My Own Hooks from Two Bugs in Claude Code Itself - DEV Community
I counted the stopped and failed tool calls in my Claude Code conversation records and had an LLM analyze the causes. About 90% of the findings worth fixing were in mechanisms I had added myself.
Ähnliche Seiten
How I stopped Claude Code from lying about finishing tasks (4 hooks, 150 lines of bash) - DEV Community