DataAI Nexus

Agentic AI / Skills

Prompt Frameworks

Skill

A reference of 30 prompt-engineering techniques with rules for choosing the cheapest accurate one — the knowledge base behind the Prompt Selector agent.

Best for

  • Anyone writing prompts regularly
  • Rebuilding a vague request into a precise one
  • Teams that want consistent prompt quality

What you give it

  • A request to rebuild, or a question like "which technique fits this?"

What you get back

  • The technique, why it fits and what it costs
  • A rebuilt prompt in that structure

How it works

  1. Holds all 30 techniques with structure, best use, cost and when to avoid each.
  2. Applies the selection rules: cheapest sufficient first.
  3. Rebuilds your prompt in the chosen structure, marking unknowns.

Example

You: Which technique for a strict JSON classification task?

Result: Few-shot (2–3 examples) for format accuracy, combined with XML-structured sections if the inputs are long. Zero-shot would be cheaper but weaker on strict formats.

Limits — please read

  • Cost and accuracy notes are rules of thumb, not measurements.
  • Techniques help; they do not add knowledge the model lacks.