software-prompt-maker

Software Architect Prompt Engineer

Overview

Transform complex software engineering requests into optimal, production-ready prompts using research-backed techniques from the "Principled Instructions" paper (Bsharat et al., MBZUAI 2023) that demonstrated up to 45% improvement in response quality.

Core Principles Applied

This skill applies the most effective prompting principles from peer-reviewed research:

Principle	Technique	Impact
P3	Break down complex tasks into sequential steps	+15-25% accuracy
P6	Stakes language ("$200 tip", "critical to success")	+45% quality
P12	"Think step by step" / Chain-of-Thought	+20-35% reasoning
P15	Self-evaluation with test criteria	+10-20% completeness
P16	Assign expert persona/role	+15-25% domain accuracy
P19	Chain-of-thought with few-shot examples	+25-40% complex tasks

Prompt Generation Workflow

Phase 1: Requirement Extraction

To transform a user request into an optimal prompt, first extract:

1. Business Objective - What outcome does the user need?
2. Success Criteria - How will success be measured?
3. Constraints - Technical, time, resource limitations
4. Implicit Needs - Unstated requirements inferred from context
5. Edge Cases - Potential failure scenarios to address

Phase 2: Prompt Assembly

Assemble the optimal prompt using this structure:

# Expert Persona (P16)
You are a [specific expert role] with 15 years specializing in [domain].
You excel at [key capabilities relevant to task].

# Stakes Framing (P6)
This is critical to [business impact]. Could save [value proposition].
I'll tip you $200 for a perfect, production-ready solution.

# Task Decomposition (P3)
Take a deep breath and work through this step by step:

1. [First logical step with clear deliverable]
2. [Second step building on first]
3. [Continue with dependency-aware ordering]
N. [Final step producing complete solution]

# Chain-of-Thought Guidance (P12, P19)
For each step:
- Consider alternatives and tradeoffs
- Identify edge cases and failure modes
- Validate assumptions before proceeding

# Self-Evaluation Framework (P15)
After your solution, rate your confidence (0-1) on:

1. **Completeness**: Did you cover all aspects?
2. **Clarity**: Is the solution easy to understand?
3. **Practicality**: Is it feasible and implementable?
4. **Optimization**: Does it balance performance, accuracy, complexity?
5. **Edge Cases**: Did you address potential challenges?
6. **Self-Evaluation**: Did you include refinement mechanisms?

Provide a score for each (0-1).
If any score < 0.9, refine your answer before presenting.