reasoning-trace-optimizer

Debug AI agents by capturing interleaved thinking traces to detect failure patterns. The tool analyzes these traces to identify where reasoning goes wrong, allowing you to iteratively optimize prompts until the agent converges on correct behavior.

The iterative prompt optimization loop consists of capture, analyze, optimize, and re-run stages. It repeatedly captures thinking traces, analyzes failure patterns, and refines prompts until the AI agent's performance converges and no longer fails.

Yes, you can generate shareable Agent Skills from prompt optimization results. After detecting failure patterns and converging on an optimized prompt, the tool outputs a shareable skill file for team use.

Yes, analyzing reasoning traces requires Python with specific dependencies. You need the anthropic, pydantic, rich, and python-dotenv packages installed in your environment to run the trace capture and optimization loops.

Fix long-horizon task failures using tool-assisted debugging on interleaved thinking traces. By analyzing the full reasoning trace, you can detect where the agent's logic breaks down over extended tasks and apply iterative prompt optimization.