Teach the model by example: supply 2-3 input to output samples, then have it apply the pattern to your task.
Force explicit step-by-step reasoning before the answer to reduce errors on harder problems.
Solve a problem several independent ways, then reconcile the attempts into one trusted answer.
Break a big, vague goal into clear, ordered, dependency-aware subtasks with done-when checks.
Design a multi-step prompt pipeline where each stage's output cleanly feeds the next.
Lock output to a strict format and schema with explicit, checkable rules and a self-check.
Steer output by naming what to avoid, each paired with the positive alternative to do instead.
Turn a rough request into one crisp, unambiguous, reusable instruction with role, tasks, and output spec.
Teaches the model a task by example so new inputs are handled in a matching style.
Forces explicit chain-of-thought reasoning before a final answer to reduce errors.
Generates multiple independent solutions and reconciles them into one trusted answer.
Breaks a large goal into ordered, verifiable subtasks with dependencies and checks.
Splits one complex prompt into a sequence of linked prompts that pass outputs forward.
Wraps a request in tight, testable constraints so output stays inside required bounds.
Specifies what the model must not do so output sidesteps known failure patterns.
Rewrites a vague prompt into a precise, role-based, structured instruction set.
Designs a balanced few-shot example set that teaches a model the exact pattern you need.
Wraps any reasoning task in explicit step-by-step structure with a separated final answer.
Breaks a large fuzzy goal into an ordered tree of independently verifiable subtasks.
Designs a multi-step prompt chain where each step's output feeds the next cleanly.
Forces a model to critique its own draft against criteria, then produce a fixed version.
Builds a scoring rubric and judging prompt to compare or grade model outputs consistently.
Diagnoses why a prompt underperforms and rewrites it through targeted, testable changes.
Generates several independent reasoning paths then reconciles them into one trusted answer.
Compiles a loose request into a structured role-context-task-constraint-format prompt.