The Anatomy of a Great Prompt: Role, Context, Task, Format

The Anatomy of a Great Prompt: Role, Context, Task, Format

Experienced prompt engineers rarely write prompts from scratch. Instead, they follow a mental template that ensures every prompt contains the right information in the right order. The most reliable framework has four components: Role, Context, Task, and Format — often abbreviated as RCTF.

Component 1: Role

The role tells the model what persona or expertise to adopt. This is not just cosmetic — it activates different areas of the model's training and shifts its vocabulary, tone, and reasoning style.

You are a senior UX copywriter specialising in SaaS onboarding flows.

You are a senior UX copywriter specialising in SaaS onboarding flows.

Effective roles are specific. "You are an expert" is weak. "You are a B2B SaaS copywriter with 10 years of experience writing onboarding emails for enterprise software" is strong because it narrows the model's frame of reference.

Component 2: Context

Context provides the background information the model needs to give a relevant answer. Without context, the model fills gaps with assumptions — and those assumptions are often wrong.

The product is a project management tool targeting remote engineering teams of 5–20 people.
The user has just completed their first project and received a completion notification.

The product is a project management tool targeting remote engineering teams of 5–20 people.
The user has just completed their first project and received a completion notification.

Good context answers: Who is the audience? What is the situation? What constraints apply? What has already happened?

Component 3: Task

The task is the actual instruction — what you want the model to do. Use a strong action verb and be precise about scope.

Write a follow-up email congratulating the user on completing their first project
and encouraging them to invite their teammates.

Write a follow-up email congratulating the user on completing their first project
and encouraging them to invite their teammates.

Weak task: "Write something about the project." Strong task: "Write a 3-sentence congratulations email with a single call-to-action to invite teammates."

Component 4: Format

Format specifies how the output should be structured. Without this, models default to whatever format they consider natural — which may not match your needs.

Output format:
Subject line: [under 50 characters]
Body: [3 short paragraphs, plain text, no markdown]
CTA: [one sentence with a hyperlink placeholder]

Output format:
Subject line: [under 50 characters]
Body: [3 short paragraphs, plain text, no markdown]
CTA: [one sentence with a hyperlink placeholder]

Putting It All Together

Role: You are a senior UX copywriter specialising in SaaS onboarding.

Context: The product is a project management tool for remote engineering teams.
The user has just completed their first project.

Task: Write a congratulations email encouraging the user to invite their teammates.

Format:
Subject: [under 50 characters]
Body: [2–3 short paragraphs, plain text]
CTA: [one sentence with a [INVITE_LINK] placeholder]

Role: You are a senior UX copywriter specialising in SaaS onboarding.

Context: The product is a project management tool for remote engineering teams.
The user has just completed their first project.

Task: Write a congratulations email encouraging the user to invite their teammates.

Format:
Subject: [under 50 characters]
Body: [2–3 short paragraphs, plain text]
CTA: [one sentence with a [INVITE_LINK] placeholder]

Using RCTF with PromptProcessor

When building batch templates in PromptProcessor, the Role, Context, and Format sections stay constant across all rows. Only the Task (or a specific variable within it) changes per row. This makes RCTF ideal for batch workflows — write the framework once, vary the input, and get consistent outputs at scale.