Prompt Engineering Sprint
Design, test, and optimize production prompts with eval loops
Takes a prompt requirement and runs a structured engineering sprint: requirement analysis, prompt design with separation of concerns, test case generation, evaluation loop, and optimization. Produces a production-ready prompt with documented edge cases.
Workflow Steps
Requirement Analysis
/prompt-creatorDefine inputs, outputs, constraints, and quality criteria
Prompt Design
/prompt-creatorDraft prompt with upstream/downstream separation and role clarity
Test Case Generation
/prompt-creatorGenerate edge cases, adversarial inputs, and golden examples
Eval Loop
/eval-loopRun prompt against test cases, score outputs, iterate on failures
Documentation
/produce-contentDocument the prompt, its constraints, and known limitations
Example
Building a production prompt for lead scoring. The sprint defines 6 input signals, designs a prompt that separates evidence gathering from scoring, generates 20 test cases (including edge cases like incomplete data), runs 3 eval iterations improving accuracy from 72% to 91%, and documents the final prompt with 4 known limitation scenarios. This is the Claude Code engineering discipline applied to AI prompts.
Related Workflows & Skills
PRD Execution
Multi-session projects that survive context loss
WorkflowProcess Documentation Capture
Turn tribal knowledge into documented, repeatable processes
SkillPrompt Creator
Systematic design for production-quality AI prompts
SkillEval Loop
Trace quality gaps to root causes, fix iteratively
SkillProduce Content
One idea into blog, newsletter, social, images
Deep Dives
Ready to automate this workflow?
Every workflow ships with Knowledge OS. Set up your system in 90 minutes.
Built and maintained by Victor Sowers at STEEPWORKS