Start with a reusable prompt
Define one task, replace variables, and make the output inspectable before asking for a more elaborate prompt.
Guides
Prompt Engineering Field Guides
Practical, source-aware methods for designing, testing, and operating reusable AI instructions.
Learning paths
Evaluate evidence and failure
Build a fixed set of cases. Distinguish missing information from bad formatting, and record actual failures rather than a confidence score.
Apply it to practical work
Follow a bounded workflow and a verification step. These teaching scenarios include authored examples, not measured production results.
No matching guides.
Editorial review
What Is a System Prompt?
Understand the persistent instruction layer that defines an AI assistant’s role, boundaries, and response behavior.
Read guide →Editorial review
System Prompt vs User Prompt
Choose the right instruction layer so reusable behavior stays separate from each individual task.
Read guide →Editorial review
How to Write Better AI Prompts
A repeatable method for turning vague requests into clear, testable AI work instructions.
Read guide →Editorial review
Zero-Shot vs Few-Shot Prompting
Know when clear instructions are enough and when examples are the fastest way to control a result.
Read guide →Editorial review
How to Get JSON Output From an LLM
Design a strict JSON contract, handle missing values, and validate model output before your application uses it.
Read guide →Editorial review
How to Reduce AI Hallucinations
Reduce unsupported claims by controlling sources, uncertainty, verification, and the scope of the task.
Read guide →Editorial review
Prompt Variables and Placeholders Explained
Turn one-off instructions into reusable templates with clear, safe, and understandable inputs.
Read guide →Editorial review
Prompt Chaining Explained
Break complex AI work into inspectable steps with explicit inputs, outputs, and quality gates.
Read guide →Editorial review
How to Test Prompt Quality
Create a small evaluation set and score prompts for accuracy, completeness, format, and practical usefulness.
Read guide →Editorial review
How to Build an AI Agent Prompt
Define an agent’s objective, tools, decision loop, stopping conditions, and escalation rules.
Read guide →Worked practice guides
Turn Meeting Notes into Action Items Without Inventing Owners
A worked meeting-note exercise with incomplete assignments, conflicting dates, a reference answer and a repeatable review checklist.
Read guide →Worked practice guides
Build an Evidence Table When Sources Disagree
Learn to separate supported facts, conflicts and unknowns using a small source packet instead of asking for an unsupported research verdict.
Read guide →Worked practice guides
Validate JSON Extraction Beyond a Well-Formed Response
A practical appointment-extraction example that separates syntax, schema, business rules and missing facts, including a local JSON exercise.
Read guide →Worked practice guides
Write a Debugging Prompt That Leads to a Reproducible Fix
Build a minimal PHP failure report, distinguish hypotheses from evidence and verify a proposed patch without exposing production secrets.
Read guide →Worked practice guides
Improve an Image Prompt by Changing One Visual Decision at a Time
An editorial image-prompt exercise focused on composition, materials and constraints, with a comparison worksheet rather than invented generation results.
Read guide →