TRACE Framework
Task · Request · Action · Context · Example
Show, don't tell. The Example field teaches the AI exactly what you want better than any description.
- Definition
- The TRACE framework (Task · Request · Action · Context · Example) is a prompt engineering structure that breaks your AI request into 5 discrete fields. It is best suited for tasks where you have an example of the ideal output.
The 5 Fields
Task
The broad category or type of task you are asking the AI to perform.
Request
The specific, precise request — what exactly you need.
Action
The concrete action the AI should take to fulfil the request.
Context
Relevant background, constraints, and situational information.
Example
A concrete example of what good output looks like. The most powerful field in TRACE.
Real Example
Scenario: Generating structured product descriptions in a specific format
Task: Product description writing. Request: Write a product description for PromptQuorum. Action: Follow the exact structure in the example. Context: B2B SaaS tool, technical audience. Example: "Notion — The all-in-one workspace. Write, plan, collaborate, and get organized. Notion is everything you need — in one tool."
When to Use TRACE
- ✓Tasks where you have an example of the ideal output
- ✓Few-shot prompting scenarios
- ✓Replicating a writing style or format
- ✓Generating structured data when you can show the schema
- ✗Tasks where no good example exists
- ✗Creative tasks where examples constrain originality
- ✗Simple factual questions (use APE)
Frequently Asked Questions
What does TRACE stand for?
TRACE stands for Task, Request, Action, Context, and Example — a few-shot framework where providing an example teaches the AI precisely what output you want.
Why is the Example field so powerful in TRACE?
Showing beats telling. A concrete example communicates format, tone, length, and style simultaneously — more efficiently than any written description.
How is TRACE different from RISEN?
TRACE uses examples to guide AI output; RISEN uses explicit sequential steps. Use TRACE when you can show what good looks like; use RISEN for ordered workflows.