Prompt workshop
What is Prompt Engineering?
Prompt engineering is designing instructions, examples, and output contracts — and, in the workshop, compiling, testing, constraining, or optimizing them.
A pasted persona belongs in the library. DSPy, promptfoo, Guidance, TextGrad, and GEPA belong in the workshop: the prompt is a program with a metric or a grammar.
Why it matters
Prompt engineering is the discipline of designing instructions, examples, and output contracts for language models. At the library level, it is writing good system prompts. At the workshop level, it is compiling, testing, constraining, or optimizing prompts as programs.
Prompts are not magic words. They are software artifacts that need versioning, testing, and maintenance — just like code.
Library vs. workshop
The library contains ready-to-use prompts: personas, templates, system instructions. They work well for many tasks and require no tooling.
The workshop is for prompts that must survive model changes, handle edge cases, or produce structured output. Workshop tools include DSPy (compilation), promptfoo (testing), Guidance (constrained decoding), and TextGrad (optimization).
Key takeaways
- 1Prompts are software artifacts — version, test, and maintain them.
- 2Library prompts are ready-to-use; workshop prompts are engineered programs.
- 3Testing prompts against assertions catches regressions early.
Common mistakes
- ✕Treating prompts as one-time text rather than maintained artifacts.
- ✕Not testing prompts against model updates.
- ✕Over-engineering prompts for simple tasks that don't need optimization.
Related terms
Concept neighborhood
Terms linked from Prompt Engineering in the glossary graph.
- Prompt Engineering
- Compiled Prompt
- Prompt Evaluation
- Context Engineering