AI / LLM Prompting Prompts

Writing, hardening, and debugging the prompts that other systems depend on. 18 prompts in this category.

• Instruction Ambiguity AuditorRun before a prompt goes into production, especially one that feeds a downstream parser.
• Few-Shot Example SelectorWhen you have more examples than context budget and need coverage rather than volume.
• System Prompt Stress TestFor any user-facing assistant before launch. Treat the outputs as regression tests.
• Output Format EnforcerWhen a parse failure downstream is more expensive than a slightly worse answer.
• Chain-of-Thought vs. Direct Answer TestUse when you are paying for reasoning tokens and are not sure they are buying anything.
• Prompt CompressionFor high-volume prompts where token cost per call actually adds up.
• Hallucination Pressure TestEssential for RAG pipelines, where the failure mode is a confident answer from missing context.
• Role Prompt CalibrationWhen the refusal behavior matters as much as the answer behavior.
• Multi-Turn Consistency CheckFor evaluating long conversations, where contradictions are the clearest memory failure signal.
• Retrieval-Augmented Prompt BuilderUse when answers need to be auditable back to a source.
• Prompt Regression Suite BuilderBefore changing a prompt that is already in production and working.
• Token Budget OptimizerWhen you are over a context or cost budget and the prompt itself is not negotiable.
• Temperature and Sampling AdvisorWhen defaults are producing too much variance or too little creativity.
• Evaluation Set GeneratorWhen you have a model but no way to tell if a prompt change made it better or worse.
• Guardrail Instruction WriterFor compliance-sensitive deployments where refusals must be consistent.
• Prompt Chaining PlannerWhen one prompt is doing extraction, transformation, and formatting all at once.
• Few-Shot Example SynthesizerWhen your few-shot set overfits to the easy cases.
• Context Window AllocatorWhen retrieval is eating your context budget and responses are getting cut off.

Browse every category →