Prompt Engineering

9 Prompt Frameworks, Explained

Every framework, with its fields, a real example, and guidance on when to use it — and when not to.

Quick Reference

Framework	Fields	Best For	Level
CO-STAR Used to win Singapore's GPT-4 competition	Context · Objective · Style · Tone · Audience · Response	Business communication and marketing copy	Medium
CRAFT	Context · Role · Action · Format · Target	Marketing copy and content creation	Medium
RISEN	Role · Instructions · Steps · End Goal · Narrowing	Enterprise workflows with sequential steps	High
TRACE	Task · Request · Action · Context · Example	Tasks where you have an example of the ideal output	High
APE	Action · Purpose · Expectation	Quick, everyday prompts	Low
SPECS	Situation · Purpose · Expected Output · Context · Style	Complex technical analysis and research tasks	High
RTF	Role · Task · Format	Professional tasks where a role matters	Low
Google Prompt	Task · Context · Persona	Business and professional tasks	Low
Single Prompt Line	One clear, direct sentence	Quick factual lookups	Low

All Frameworks

CO-STAR

Medium

Context · Objective · Style · Tone · Audience · Response

Credited to GovTech Singapore's Data Science & AI team, and used to win Singapore's first GPT-4 prompt engineering competition. Maximum control over voice, audience, and output format.

CRAFT

Medium

Context · Role · Action · Format · Target

Role-based framing with a clear target audience. Great for marketing, copywriting, and creative work.

RISEN

High

Role · Instructions · Steps · End Goal · Narrowing

Built for multi-step enterprise tasks. The Narrowing field keeps AI output on-track and within constraints.

TRACE

High

Task · Request · Action · Context · Example

Show, don't tell. The Example field teaches the AI exactly what you want better than any description.

APE

Low

Action · Purpose · Expectation

Three fields, high impact. The fastest framework — ideal for everyday prompts before graduating to more complex structures.

SPECS

High

Situation · Purpose · Expected Output · Context · Style

Detail-oriented and precise. The Expected Output field eliminates guesswork and is ideal for complex technical tasks.

RTF

Low

Role · Task · Format

Minimum viable prompt structure. Three fields that cover the three most impactful variables in any prompt.

Google Prompt

Low

Task · Context · Persona

Developed and recommended by Google. Optimised for Gemini and professional workplace tasks.

Single Prompt Line

Low

One clear, direct sentence

No framework fields — just a single well-crafted sentence. Faster to write, still effective for simple queries.