skill-review

Prompt compliance analysis reads prompt files and repository history to verify model execution accuracy and implementation correctness. It uses natural language processing and code comparison techniques to evaluate the quality of the dialogue process and output evaluation results.

Source code mode analyzes code changes and compares them with prompts to ensure reasonable implementation, while dialogue mode focuses on the quality of the dialogue process by extracting conclusions from model responses. Both modes evaluate prompt compliance but target different aspects of execution results.

Yes, git and python are required dependencies to perform prompt compliance analysis. Git provides access to repository history for analyzing code changes, while python executes the scripts needed to run the implementation review and generate evaluation files.

Dialogue mode evaluates dialogue process quality by extracting conclusions directly from model responses. It strictly follows natural language output principles to assess whether the conversation meets prompt requirements, focusing on the interaction quality rather than underlying code changes.

The best way to verify implementation correctness is using source code mode, which analyzes repository code changes and compares them directly with prompt files. This ensures the implementation is reasonable and meets all specified requirements through code comparison techniques.

You should not use natural language output when you need machine-readable metrics, as this skill strictly follows natural language output principles for evaluation results. It focuses on qualitative assessment of the conversation process rather than generating structured data formats.