1. Identify the work to evaluate from conversation history or user input.
2. Extract relevant context: original task, output, files involved, and evaluation focus.
3. Provide the evaluation scope to the user for clarity.
4. Launch a judge sub-agent with a tailored prompt and evaluation criteria.
5. Validate the judge's evaluation for accuracy and completeness.
6. Present the evaluation report to the user with key findings and recommendations.
7. Offer follow-up options: address improvements, request clarification, or proceed as-is.
SynapticSkills