system-prompt-archaeology

To reverse-engineer behavioral patterns from an AI system prompt, you need to decompose the text into layered categories like identity, safety, and rules, then map those signals to behavioral dimensions such as pacing, depth, and tone.

Extracting a structured disposition preset from a prompt involves classifying behavioral signals across dimensions like initiative, validation, and error-handling, then translating those mapped patterns into a structured YAML format for reuse.

Yes, you can compare prompt patterns across different AI tools by decomposing each system prompt into layered behavioral classifications and generating standardized disposition presets to identify structural differences in agent behavior.

You generate a YAML disposition preset from a system prompt by analyzing the text for behavioral rules and tool definitions, classifying those signals into known behavioral clusters, and outputting the result as a YAML file.

Optional file-system tools are required to store the generated disposition preset locally, allowing you to save the extracted YAML output directly to your environment for downstream integration and analysis.

When decomposing a system prompt, the analysis breaks down the text into distinct layers including identity, safety constraints, tool definitions, and behavioral rules to map out the agent's underlying architecture.