Writing Effective Prompts
Master the art of prompt engineering to get better, more accurate responses from any AI model on Agentyk.me.
Why Prompts Matter
The quality of your AI output depends heavily on how you phrase your input. A well-crafted prompt can mean the difference between a generic response and a highly useful, tailored answer.
The CRAFT Framework
Use the CRAFT framework to structure prompts that consistently produce great results:
- 1
Context
Provide background information. Who are you? What is the project? What has already been done?
- 2
Role
Tell the AI what role to adopt — e.g., "Act as a senior backend developer reviewing a pull request."
- 3
Action
Clearly state what you want the AI to do. Use action verbs: write, summarize, compare, debug, refactor.
- 4
Format
Specify the output format: bullet list, JSON, table, code block, step-by-step guide, etc.
- 5
Tone
Set the tone if it matters: formal, casual, technical, beginner-friendly.
Example: Before & After
Weak prompt
Write me a landing page.Strong prompt
Act as a senior front-end developer. Write the HTML and Tailwind CSS for a SaaS landing page for an AI writing tool aimed at content marketers. Include a hero section with headline, subheadline, CTA button, a 3-column feature grid, and a testimonial section. Use a clean, modern style with a blue/white color scheme.Common Mistakes to Avoid
- Being too vague — "Help me with code" gives the AI nothing to work with.
- Overloading a single prompt — break complex tasks into smaller steps.
- Forgetting to specify format — the AI will guess, and it might guess wrong.
- Not iterating — treat AI conversations like a collaboration, refine as you go.
Quick Reference Cheat Sheet
- Start with context and role.
- State the action clearly with a verb.
- Specify the output format.
- Add constraints (word count, tone, language).
- Iterate on the result — ask follow-up questions.
