reasoning-trace-optimizer

Debug AI agents by capturing detailed reasoning traces to identify failure patterns like context degradation. The Skill analyzes interleaved thinking blocks and tool calls to pinpoint exactly why an agent is failing, enabling targeted prompt optimization.

Reasoning trace analysis detects instruction drift by recording detailed thinking blocks and tool calls during agent interactions. It identifies common failure patterns where the agent loses track of original instructions, providing data to refine and optimize prompts.

You can optimize prompts automatically through an iterative optimization loop. The Skill analyzes captured reasoning traces using interleaved thinking capabilities to generate improved prompts that enhance agent performance and resolve identified issues.

Fix tool confusion by analyzing AI agent reasoning traces to detect when agents misuse or misunderstand available tools. The Skill identifies these specific failure patterns and suggests prompt improvements to clarify tool usage and workflow execution.

You need a Python environment with dependencies including anthropic, pydantic, rich, and python-dotenv. These libraries support trace capture, pattern analysis, and the iterative optimization loop required to analyze and refine agent reasoning.