agentsop-prompt-history-inspect

What problem does it solve?

This Skill prevents wasted debugging by ensuring you inspect what the model actually received—because the prompt you wrote is often not the prompt that was rendered and sent.

Core Features & Use Cases

• 30-second first-move inspection: forces a brief inspect step before editing prompts, swapping models, retrying, or changing generation settings.
• Cross-framework prompt/LLM-input visibility: provides a unified SOP and mapping for DSPy, LangChain, LangGraph, CrewAI, Aider, and raw OpenAI/Anthropic SDK logging.
• Safer, faster root-cause classification: helps distinguish whether the divergence is in your code, the framework render layer, or the provider transport layer.
• When to use: ideal for wrong answers, schema/format failures, truncation/refusals, latency/cost spikes, and production bugs that are hard to reproduce.

Quick Start

Ask your agent to load agentsop-prompt-history-inspect the moment an LM call output is surprising, then dump the rendered prompt/LM input before making any other debug change.