Home Framework Library RACE

RACE

Role-Action-Context-Expectation

The framework for work with a defined actor doing a defined job.

Product & Growth Free Framework 4 Steps

What is RACE?

RACE splits a prompt into four things: Role, Action, Context and Expectation. It is the framework to reach for when there is a specific person doing a specific job — a QA engineer writing a bug report, a PM writing a spec — and the perspective matters more than the voice.

The reason RACE works is the slot most other frameworks put last, or leave out. Role comes first, and a role is not decoration: "you are a senior QA engineer writing for the backend team" changes what the model assumes the reader already knows, which vocabulary needs no explanation, and what would be embarrassing to state. Product and engineering writing lives or dies on that calibration — the same facts written for a stakeholder and for the team that owns the code are two different documents.

Best used for
User stories product requirements UX specifications feature briefs

Where RACE Came From

A modern acronym, not a borrowed one

Unlike AIDA or STAR, RACE has no single documented author and no pre-AI history. It is one of a family of role-first prompt structures that appeared as people worked out, largely in public and largely by trial, that telling a model who it was changed its output more than telling it to try harder. Anyone claiming a definitive inventor is guessing, and we are not going to.

What it inherits

Its lineage is clearer than its authorship. RACE is the prompt-engineering descendant of the work brief — the form used to hand a task to a colleague, which has always opened by establishing who is doing it and for whom. Read the four slots as a brief rather than as a formula and the order stops looking arbitrary.

Why it lands in product work

It suits product and engineering because those disciplines already write this way. A user story is literally Role plus Action plus Context ("as a returning customer, I want to reorder a previous purchase, so that I do not re-enter my details"). RACE is that shape with an explicit Expectation bolted on, which is the half user stories habitually leave to a conversation.

The 4 Slots, One at a Time

Each slot is a decision. Leave it out and the model still makes it — just without you.

R
Role
Who is writing, and for whom. Both halves matter.

The slot that does the most work in RACE, and the one people fill most lazily. "You are a developer" is barely narrower than nobody. A role earns its place when it fixes seniority, discipline and audience at once, because that combination is what sets the level of explanation.

Weak You are a developer.
Strong You are a senior QA engineer writing for the backend team that owns authentication.
A
Action
The single deliverable, stated as a verb plus an artefact.

Action is deliberately narrow: one thing. If your Action contains an "and", you probably have two prompts. The commonest failure is naming a topic rather than an artefact — "cover the login problem" leaves the model to decide between a bug report, an incident summary and a customer email.

Weak Cover the login problem.
Strong Write a bug report for an intermittent login failure.
C
Context
The facts the model cannot infer, and the ones it must not invent.

In product work Context is mostly evidence: what was observed, when it started, what is already ruled out. This is also the slot where you say what you do NOT have — stating "we have request IDs but no stack traces" is what stops a model producing a confident trace it made up.

Weak Users are having trouble logging in sometimes.
Strong Roughly one in twenty sign-in attempts returns a 500 on /auth/session. It started after Tuesday’s deploy, affects only the legacy SSO path, and does not reproduce on staging. We have request IDs but no stack traces.
E
Expectation
What finished looks like: shape, length, and the quality bar.

RACE has four slots, so Expectation has to carry what CO-STAR spreads across Style, Tone and Response. Treat it as three questions rather than one — what shape, how long, and what would make this wrong. The third is the one that separates a usable Expectation from a decorative one.

Weak Make it clear and well structured.
Strong Under 300 words using exactly these seven headed sections: Summary, Environment, Steps to reproduce, Expected, Actual, Evidence, Suspected cause. Label every claim as observed or inferred. Do not invent log lines we do not have.

One Task, Before and After

The task: File a bug report for an intermittent login failure. Both prompts below are scored by our free prompt checker — paste either in and you will get the same number, because the scoring is deterministic.

30 Without RACE · weak
Write a bug report for the login issue.

Eight words. The model has to invent the environment, the symptoms, the severity and the format — and its guesses will read as plausible, which is worse than reading as wrong.

86 With RACE · strong
Role: You are a senior QA engineer writing for the backend team that owns authentication.

Action: Write a bug report for an intermittent login failure.

Context: Roughly one in twenty sign-in attempts returns a 500 on the /auth/session endpoint. It started after Tuesday's deploy, affects only users on the legacy SSO path, and does not reproduce on staging. We have request IDs but no stack traces. The team reading this has not seen the incident channel.

Expectation: Under 300 words, using exactly these seven headed sections - Summary, Environment, Steps to reproduce, Expected, Actual, Evidence, Suspected cause. Work through the evidence before proposing a cause. Label every claim as either observed or inferred, for example 'Observed: 4 of 80 requests returned 500'. Do not guess at a root cause the evidence does not support, and do not invent log lines or stack traces we do not have.

One hundred and forty-four words, and the highest-scoring worked example on any of these framework pages. Four slots, used properly, beat six used lazily.

Why RACE scores higher than most frameworks here

At 86 this is the strongest worked example across the framework pages, and the reason is instructive: RACE has fewer slots than CO-STAR but its Expectation slot is an open container. Anything you can specify about the finished artefact goes there, so a well-used RACE prompt picks up points a six-slot framework has no room for:

Reasoning8 / 8

"Work through the evidence before proposing a cause" fits naturally inside Expectation. CO-STAR has nowhere to put it.

Exclusions6 / 6

"Do not invent log lines we do not have" is the single most valuable line in a technical prompt, and Expectation absorbs it.

Format12 / 12

Seven named sections with a word cap. Numbers in the Expectation slot are what make it unmisreadable.

The lesson is not that RACE is the best framework. It is that a small framework with one open slot often beats a large one with six closed slots, because the open slot lets you say the specific thing your task actually needs. The risk is the mirror image: an Expectation left as "make it clear" wastes the only flexible slot RACE has, and the prompt collapses to about 45.

Copy-Paste Prompt Template

Replace the bracketed placeholders with your specific details.

Role: [Who is writing, and who reads it — seniority, discipline, and what the reader already knows]

Action: [One deliverable: a verb plus an artefact. If it contains 'and', split it]

Context: [The facts the model cannot infer — including what you have ruled out, and what you do NOT have]

Expectation: [Shape with numbers in it, length, and one line about what would make the answer wrong]

When RACE Fits — and When It Does Not

Reach for it
  • Product and engineering documents: specs, bug reports, release notes, technical briefs.
  • Anything where the reader’s existing knowledge decides the level of explanation.
  • Writing as a role you do not personally hold — a PM drafting for engineers, or the reverse.
  • Tasks with a clear finished artefact you can describe in numbers.
Use something else
  • Work where voice and emotion matter more than perspective. Use CO-STAR — RACE has no tone slot.
  • Persuasion. There is no place in RACE for a reader’s emotional journey. Use AIDA or PAS.
  • Open exploration. RACE assumes you know the artefact; if you do not, you are using it too early.
  • Very short asks. Four labelled slots for a one-line Slack message is more scaffolding than message.

10 Ready-Made RACE Prompts

Every prompt below was produced by the Frompting generator with RACE selected — not written by hand for this page. Each is scored by our prompt checker; the median is 82/100. Click one to open it, then copy.

Product requirements document 78
You are a product manager tasked with creating a comprehensive product requirements document for a new feature.

Draft the document for a feature called **[FEATURE_NAME: brief descriptive title]** aimed at **[TARGET_USER: primary user persona or segment]**. Include the business objective this feature should achieve, such as increasing engagement, reducing friction, or generating revenue, and note any key constraints (e.g., technical limits, regulatory requirements, budget caps) as **[CONSTRAINTS: list of critical constraints]**.

The document must contain:
- A clear description of the problem the feature solves and the user need it addresses.
- Detailed functional requirements, specifying user actions, system responses, and any UI/UX considerations.
- Non‑functional requirements covering performance, security, accessibility, and scalability.
- Acceptance criteria that define when the feature is complete and meets the intended outcome.
- Success metrics to evaluate the feature’s impact after launch.

Write the requirements in a structured markdown format with headings for each section, using concise bullet points where appropriate. Keep the total length between 800 and 1,200 words. Ensure the language is precise, avoids ambiguity, and aligns with standard product management best practices. Do not include any speculative business claims; if any detail is unknown, mark it with a placeholder in square brackets.
205 words · scores 78/100 solid
Release notes for an app update 76
You are a product communication specialist tasked with drafting concise release notes for a mobile app update.

Write clear, user‑focused release notes that list the new version number, the app name, and each highlighted change. Use a brief introductory sentence, followed by a bulleted list of improvements, new features, bug fixes, and any known issues. Keep the total length between 120 and 180 words.

[APP_NAME: supply the name of the mobile app]
[VERSION_NUMBER: supply the new version identifier]
[KEY_CHANGES: list the main new features, enhancements, and bug fixes]
[KNOWN_ISSUES: list any remaining issues to be disclosed, if any]

If any of the placeholders above are unknown, state your assumption and proceed, or ask up to three clarifying questions before finalizing the notes.

**Quality criteria**
1. Each bullet starts with an action verb and is no longer than one sentence.
2. Language is plain, avoids technical jargon, and highlights user benefit.
3. No promotional slogans or marketing copy; focus solely on factual update information.

**Boundary** – Do not include pricing details, future roadmap items, or internal development notes.

Write this for [AUDIENCE: who will read the output, and how much they already know]. Match the depth, vocabulary and examples to that reader.
200 words · scores 76/100 solid
User onboarding email sequence 86
You are a product‑focused copywriter who specializes in crafting concise, engaging onboarding email sequences that guide new users toward their first “aha” moment.

Write a series of three onboarding emails for a SaaS product that welcomes new sign‑ups, introduces the core value proposition, and encourages the first key action that demonstrates the product’s benefit.

The emails will be sent to **[AUDIENCE: brief description of the target user segment, e.g., “new users who have just created an account”]**.

Each email should be no more than **150 words** and follow a clear, friendly tone that balances persuasive language with helpful guidance.

Include a compelling subject line for each email, a brief preview text, and a single clear call‑to‑action button label.

Quality criteria:
1. The sequence should progressively build familiarity, moving from greeting to value demonstration to activation.
2. Messaging must align with the product’s core job‑to‑be‑done (functional, emotional, and social aspects) and highlight the primary “aha” moment.
3. Language should be natural and varied, avoiding repetitive phrasing or overly formal constructions.

Exclude any mention of pricing, legal terms, or unrelated product features.

If any of the above details are unclear, state your assumptions and ask up to three clarifying questions before drafting the emails.
208 words · scores 86/100 strong
Bug report for a dev team 82
You are a meticulous QA engineer tasked with documenting software defects for a development team.

Draft a clear, concise bug report that captures all essential details so developers can reproduce and fix the issue efficiently.

The report should pertain to the following context: the defect occurs in **[PRODUCT_NAME]** version **[VERSION]**, observed on **[OPERATING_SYSTEM / BROWSER / DEVICE]**. Include a brief summary, steps to reproduce, expected behavior, actual behavior, severity level (**[SEVERITY]**), any relevant screenshots or logs, and the environment configuration (**[ENVIRONMENT_DETAILS]**).

Your output must be a markdown‑formatted bug report no longer than 250 words, organized with headings for each section.

Quality criteria:
- Completeness: all required fields are present.
- Clarity: steps are numbered and easy to follow.
- Precision: severity and environment are accurately described.

Exclude any speculation about root cause or suggested fixes; focus solely on factual observation. If any required detail is unknown, indicate it with a placeholder in the form **[PLACEHOLDER: description]**.

Before writing the final answer, work through the problem step by step and weigh the main trade-offs; present only the reasoned conclusion, not your working notes.
183 words · scores 82/100 solid
Technical spec for an API integration 80
You are an experienced API architect tasked with producing a comprehensive technical specification for an API integration.

Draft a detailed specification that outlines the integration’s purpose, functional endpoints, data models, authentication mechanisms, error handling, rate limiting, versioning strategy, and security considerations.

The integration will connect **[SYSTEM_A]** with **[SYSTEM_B]**, using **[PROTOCOL]** (e.g., REST, GraphQL, gRPC) and **[AUTH_METHOD]** (e.g., OAuth 2.0, API keys). Include any required request/response schemas, sample payloads, and a concise diagram description of the data flow. Identify any dependencies on external services or libraries and note deployment environment constraints.

Deliver the specification in markdown format, organized with clear headings for Overview, Endpoints, Data Models, Authentication, Error Handling, Rate Limiting, Versioning, Security, and Deployment. Keep the total length between 1,200 and 1,500 words. Ensure the document is technically precise, uses appropriate terminology, and provides brief inline definitions for any specialized terms.

Quality criteria:
- Completeness: all required sections are present and logically ordered.
- Clarity: definitions, schemas, and examples are easy to understand for developers.
- Accuracy: technical details align with the chosen protocol and authentication method.

Exclude any marketing language, pricing information, or unrelated business claims. If any critical details are missing, state your assumptions explicitly and ask up to three clarifying questions before finalizing the specification.
212 words · scores 80/100 solid
Support macro: password reset 89
You are a customer support specialist tasked with creating a concise, step‑by‑step macro that agents can follow to reset a user’s password.

Draft the macro as a numbered list, using clear, direct language suitable for copy‑and‑paste into a support ticketing system. Each step should include any required placeholders (e.g., [USER_EMAIL], [RESET_LINK]) and brief instructions on how to personalize the response. Keep the entire macro under 150 words.

Assume the macro will be used by agents handling inbound requests via live chat and email. The tone should be friendly yet professional, reflecting the brand’s voice. If any platform‑specific details (such as the name of the password‑reset portal or security verification method) are unknown, insert a placeholder in the format [PLACEHOLDER: description] and do not fabricate information.

The final output must be a ready‑to‑use macro that agents can copy verbatim, with no extra commentary or headings. Ensure the steps are logically ordered, cover verification of identity, provision of the reset link, and a brief closing offering further assistance.

If any bracketed detail above is left unfilled, choose a sensible value from the context, state that assumption in one line before you begin, and continue — do not ask for it and stop.
211 words · scores 89/100 strong
Sprint retrospective summary 82
You are a skilled agile facilitator tasked with drafting a concise sprint retrospective summary.

Summarize the recent sprint by first outlining the sprint’s duration, primary objectives, and any notable outcomes or challenges. Then capture the team’s reflections on what went well, what didn’t, and actionable improvement ideas. Structure the summary for easy reading by the development team and stakeholders.

[TEAM NAME]: the name of the team the summary is for.
[SPRINT NUMBER OR DATE RANGE]: the identifier or dates of the sprint.
[KEY METRICS OR GOALS]: the main goals or metrics the team aimed to achieve.
[PRIMARY AUDIENCE]: who will read this summary (e.g., developers, product owners, executives).

Deliver the summary in 200–300 words, using clear headings for “What Went Well,” “What Could Be Improved,” and “Action Items.” Ensure the tone is constructive and forward‑looking, and highlight at least two specific action items with owners and target dates. Include a brief opening paragraph that sets the context and a closing line that reinforces the team’s commitment to continuous improvement.

If any bracketed detail above is left unfilled, choose a sensible value from the context, state that assumption in one line before you begin, and continue — do not ask for it and stop.
207 words · scores 82/100 solid
Feature announcement: in-app messaging 74
You are a product marketing copywriter tasked with crafting a concise feature announcement for a new in‑app messaging capability.

Write a compelling announcement that introduces the feature, highlights its primary benefits, and encourages users to start using it immediately. Tailor the tone to be enthusiastic yet professional, suitable for a tech‑savvy audience familiar with the product.

The announcement will be sent to [AUDIENCE: specify the target user segment, e.g., “existing app users” or “beta testers”] and will appear in the app’s news feed and email newsletter. Assume the feature will launch on [LAUNCH DATE: provide the planned release date] and is named [FEATURE NAME: supply the official name of the in‑app messaging feature]. Emphasize at least two key advantages (e.g., real‑time communication, seamless integration, enhanced engagement) and include a clear call‑to‑action prompting users to try the feature now.

Deliver the copy in plain text, no longer than 180 words. Ensure it:
- Clearly conveys the new capability and its value.
- Uses active, benefit‑focused language.
- Ends with a direct, motivating call‑to‑action.

State any assumptions you make and ask up to three clarifying questions if needed before finalizing the announcement.
201 words · scores 74/100 solid
Competitor analysis brief 95
You are a seasoned product strategist tasked with creating a concise competitor analysis brief for a product team.

Develop a brief that:
- Summarizes the competitive landscape, highlighting the most relevant rivals and their positioning.
- Evaluates each competitor across key dimensions (market share, pricing, feature set, go‑to‑market strategy, strengths and weaknesses).
- Identifies strategic gaps and opportunities for the product team to leverage.

The brief is for: [AUDIENCE: specify the product team’s role, e.g., “product managers and designers”].

Include the following sections, each no more than 150 words:
1. **Overview** – one‑paragraph snapshot of the market context and why the analysis matters.
2. **Competitor Profiles** – a table with columns: Competitor, Core Offering, Pricing Model, Market Share (approx.), Key Strengths, Key Weaknesses.
3. **Comparative Insights** – bullet points (max 6) that contrast the focal product with the competitors on critical factors.
4. **Strategic Recommendations** – up to three actionable suggestions for the product team, grounded in the analysis.

Quality criteria:
- Use recent, verifiable industry data where available; mark any estimates as “approx.”.
- Apply Porter’s Five Forces concepts to frame the competitive pressures without naming the model.
- Keep language clear, business‑focused, and free of unnecessary jargon.

Boundary: Do not include detailed financial forecasts or internal company confidential information.

If any required details are unknown (e.g., product name, primary market, list of top competitors, analysis timeframe), state your assumptions and ask up to three clarifying questions before finalizing the brief.

Before writing the final answer, work through the problem step by step and weigh the main trade-offs; present only the reasoned conclusion, not your working notes.

Use this table shape, one row per item, filling values from your analysis:

| Competitor | Core Offering | Pricing Model | Market Share |
| --- | --- | --- | --- |
290 words · scores 95/100 strong
User research interview guide 89
You are a seasoned UX researcher tasked with creating a comprehensive user research interview guide.

The guide must be organized to capture the target user’s role, the specific actions they perform, the context in which those actions occur, and the outcomes you expect to learn about. Structure the interview questions so they flow naturally from understanding who the participant is, through what they do, the circumstances surrounding those actions, and finally the insights you aim to gather.

Include the following sections, each limited to the indicated length:

1. **Participant Profile** - a brief description of the user segment you will interview (up to 80 words).
2. **Research Objectives** - list 3-5 clear goals for the interview (up to 60 words).
3. **Interview Script** - a sequenced set of open-ended questions covering role, actions, context, and expected outcomes (approximately 12-15 questions, total ≤ 250 words).
4. **Logistics** - recommended interview duration, setting, and any required materials (up to 40 words).

Quality criteria:
- Questions are open-ended, unbiased, and encourage detailed storytelling.
- The flow logically progresses from role to outcomes without jumping between topics.
- Language is clear, jargon-free, and suitable for participants with varying familiarity with the product.

Boundary: Do not include any demographic screening criteria or recruitment scripts; focus solely on the interview guide content.

If any of the following details are unknown, insert a placeholder in the format **[PLACEHOLDER: brief hint]**:
- **[USER_SEGMENT: description of the target users]**
- **[PRODUCT_TYPE: type of product or service being studied]**
- **[INTERVIEW_LENGTH: desired length of each interview]**
- **[NUMBER_OF_PARTICIPANTS: planned sample size]**

State any assumptions you make and ask up to three clarifying questions before finalizing the guide.

Note on length: this supersedes any word count given above. Covering the 4 sections needs roughly 535–705 words in total. Use that range rather than compressing any section to fit a smaller one.

Before writing the final answer, work through the problem step by step and weigh the main trade-offs; present only the reasoned conclusion, not your working notes.
331 words · scores 89/100 strong

Scores range from 74 to 95. They are shown as generated rather than cherry-picked — a library where every entry scores in the nineties tells you it was curated, not measured.

RACE vs the Alternatives

CO-STAR Context, Objective, Style, Tone, Audience, Response

Swap to CO-STAR when the output is read for its voice rather than its content — a post, a newsletter, an announcement. It has Style and Tone, which RACE does not.

RTF Role, Task, Format

RACE minus Context. Right when the model needs no background at all — a pure transformation like reformatting or extraction, where the input carries everything.

TAG Task, Audience, Guardrails

Use TAG when the constraints matter more than the perspective. Its Guardrails slot is explicit where RACE buries exclusions inside Expectation.

ICIO Instruction, Context, Input, Output

Better when there is material to work ON. ICIO has a dedicated Input slot; in RACE that content has to be smuggled into Context, where it competes with the background.

APE Action, Purpose, Expectation

The faster cousin. Drop Role and add Purpose when the job is a short internal message and the reason matters more than the perspective.

Five Ways People Get RACE Wrong

1
Writing a role instead of a perspective

"You are a helpful assistant" and "you are a developer" both score close to nothing. The role has to answer who is writing and for whom. Half of RACE’s value is in that second half, and most people omit it entirely.

2
Putting two deliverables in Action

"Write the bug report and draft the customer update" produces a compromise between two documents. Run it twice with the same Role and Context instead.

3
Using Context as a product description

Context is the evidence for this task. Company background the model does not need crowds out the detail it does — and the detail it does need is usually the awkward one, like what you have already ruled out.

4
Leaving Expectation vague

The gravest RACE-specific error, because Expectation is its only flexible slot. "Clear and well structured" wastes it. Numbers, named sections, and one line about what would make the answer wrong.

5
Forgetting to say what you do not have

Technical prompts invite confident invention: stack traces, log lines, version numbers. Naming the gap in Context is what stops it.

RACE Questions

What does RACE stand for?

Role, Action, Context, Expectation. Four slots, with Role covering both who writes and who reads.

Who invented the RACE framework?

Nobody documented. It is a modern prompt-engineering acronym with no single author, unlike AIDA or STAR which predate AI by decades. Its shape is inherited from the ordinary work brief, and from the user story in particular.

Is RACE the same as the RACE marketing model?

No, and the collision causes real confusion. The marketing RACE is Reach, Act, Convert, Engage — a digital marketing planning model, unrelated to prompting. If a search result is talking about customer funnels, it is the other one.

Should Role go first?

Yes, and the acronym already puts it there. Models weight the opening of a prompt most heavily, and the role colours how everything after it is read. A role stated at the end is noticeably weaker than the same role stated at the start.

Can I use RACE for creative writing?

You can, and it will feel like a brief rather than an invitation. Creative work usually wants tone specified, which RACE has no slot for. CO-STAR is the better fit.

What if I do not have context to give?

Then you want RTF, which is RACE without the Context slot. An empty Context is worse than no Context slot, because the model reads the gap as "there is nothing relevant here".

Generate a RACE Prompt Instantly

Skip the manual template — Frompting applies RACE to your topic in one click.

Try it Free

Framework Details

Name RACE
Stands for Role-Action-Context-Expectation
Domain Product & Growth
Steps 4
Access Free
Steps at a glance
R Role
A Action
C Context
E Expectation