Task 4.1
Turn quality preferences into testable criteria
Design prompts with explicit criteria to improve precision and reduce false positives
Open deckStudy decks / Domain 4
Prompt Engineering & Structured Output
Precision criteria, few-shot prompting, JSON schemas via tool use, validation loops, and batching.
Complete domain coverage: 6 of 6 lesson decks, each mapped to the lesson objectives.
Task 4.2
Teach judgment with a few boundary examples
Apply few-shot prompting to improve output consistency and quality
Open deckTask 4.3
Guarantee structure without confusing it for truth
Enforce structured output using tool use and JSON schemas
Open deckTask 4.4
Validate meaning, then retry only what can change
Implement validation, retry, and feedback loops for extraction
Open deckTask 4.5
Batch only work whose deadline tolerates uncertainty
Design efficient batch processing strategies
Open deckTask 4.6
Review with independent context and multiple passes
Design multi-instance and multi-pass review architectures
Open deck