Prompt Optimizer (EARS)

Goal

Transform fuzzy asks into clear, testable requirements and a production-ready prompt, using EARS syntax with domain theory support.

Guardrails

• Keep statements behaviorally testable; prefer measurable thresholds over adjectives.
• One action per EARS statement; avoid compound sentences.
• Include edge/error cases alongside happy paths.
• Do not invent domain policies; ground in recognized frameworks only.
• Examples must use realistic data (no placeholder lorem).

Workflow

1. Assess the ask.
   • Note missing triggers, actors, constraints, metrics.
   • Identify domain (productivity, UX, security, etc.).
2. Draft EARS statements.
   • Use patterns: Ubiquitous, Event-driven, State-driven, Conditional, Unwanted-behavior.
   • Expand compound requirements into atomic items.
3. Ground with domain theories.
   • Pick 2–4 relevant frameworks from references.
   • Map each to specific behaviors/constraints (e.g., Hick’s Law → limit choices, BJ Fogg → prompt + ability + motivation).
4. Add concrete examples.
   • Provide success, failure, and edge scenarios with realistic inputs/outputs.
5. Assemble the enhanced prompt.
   • Structure with Role, Skills, Workflows, Examples, Formats.
   • Call out required inputs and expected outputs clearly.
6. Present results.
   • Show issues found, EARS list, chosen theories, final prompt, and usage guidance.

Output Template

## Original Requirement
<user text>

**Issues:** <bulleted gaps/ambiguities>

## EARS Transformation
1. ...

## Domain & Theories
- <Theory> – <application>

## Enhanced Prompt
# Role
...
## Skills
- ...
## Workflows
1. ...
## Examples
- Success: ...
- Failure: ...
- Edge: ...
## Formats
- Output shape, file types, style/constraints

**How to use:** <brief guidance>

References (load when needed)

• prompt-optimizer/references/ears_syntax.md: full EARS patterns and rewrite tips.
• prompt-optimizer/references/domain_theories.md: theory catalog by domain.
• prompt-optimizer/references/examples.md: complete before/after transformations.
• prompt-optimizer/references/advanced_techniques.md: multi-stakeholder, non-functional, complex logic.