Skip to main content

Documentation Index

Fetch the complete documentation index at: https://mintlify.com/nidhinjs/prompt-master/llms.txt

Use this file to discover all available pages before exploring further.

The six text templates cover the full spectrum of language-model tasks: quick one-shot requests, polished professional documents, multi-step projects, creative brand work, logic and debugging problems, and pattern-replication tasks where format consistency matters more than instruction length. Prompt Master picks exactly one template per request — never a blend — and populates every required component from your input before running the token-efficiency audit.
All six templates target LLMs (Claude, GPT-4o, Gemini, etc.) and any tool that accepts a freeform text prompt. For coding agents and image generators, see Agentic Templates and Image Templates.

​
RTF — Role, Task, Format

RTF is the fastest template in the library. It strips a request down to three load-bearing components and nothing else. Prompt Master routes to RTF when the task is a single-shot operation with a clear deliverable and no need for stylistic nuance or multi-step reasoning. Routed when: the task is direct, the output format is obvious, and the request can be completed in one model turn without follow-up.
ComponentWhat it carries
RoleA domain-specific expert identity that sharpens the model’s response register
TaskThe single, concrete action the model must perform
FormatThe exact output structure — bullet list, table, numbered steps, JSON, etc.
Example — Prompt Master output using RTF: 
You are a senior technical writer who specializes in API documentation.

Task: Write a one-paragraph description of the POST /auth/token endpoint for a REST API reference doc. The endpoint accepts a client_id and client_secret in the request body and returns a short-lived JWT.

Format: A single paragraph of 3–4 sentences. Start with what the endpoint does. End with the token expiry behavior. No code blocks. No headings.

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

CO-STAR is the text template for work that will be read by a human audience in a professional context. It adds three dimensions that RTF omits — style, tone, and audience — which are the difference between a technically correct document and one that actually lands with its readers. Routed when: the task involves a professional document, report, business email, proposal, press release, or any deliverable where voice and audience shape the output as much as content does.
ComponentWhat it carries
ContextBackground the model needs to understand the situation
ObjectiveThe specific outcome the document must achieve
StyleWriting register — formal, conversational, technical, journalistic
ToneEmotional register — authoritative, empathetic, urgent, reassuring
AudienceWho will read this — their role, expertise level, and what they care about
ResponseFormat, length, structure, and any section-level requirements
Example — Prompt Master output using CO-STAR: 
Context: Acme Corp is launching a new B2B SaaS product called FlowSync, a workflow automation platform targeting mid-market operations teams. The launch is in 3 weeks. This email is going to existing customers who currently use the company's legacy product, FlowLite.

Objective: Convince existing FlowLite customers to attend a 30-minute live demo of FlowSync and join the early-access waitlist.

Style: Professional but energetic. Concrete over abstract — use specific feature names and outcomes rather than generic benefit language.

Tone: Confident and generous. The reader should feel like they're being given early access to something valuable, not sold to.

Audience: Operations managers and team leads at companies with 100–500 employees. They are busy, skeptical of marketing copy, and care about ROI and ease of migration.

Response: A plain-text email with a subject line, 3 short body paragraphs (max 60 words each), and a single CTA button label. No bullet lists. No headers inside the body.

​
RISEN — Role, Instructions, Steps, End Goal, Narrowing

RISEN is the template for tasks that are too complex for a single instruction block. It breaks the work into an ordered sequence of steps and adds an explicit narrowing component — constraints that prevent the model from drifting into adjacent territory as it works through a long, multi-part task. Routed when: the task has multiple sequential phases, requires the model to make decisions across several distinct subtasks, or has clear boundaries that must not be crossed during execution.
ComponentWhat it carries
RoleExpert identity calibrated to the full scope of the project
InstructionsHigh-level directive — what the overall task is
StepsOrdered list of discrete phases the model must work through
End GoalThe final deliverable — what “done” looks like
NarrowingExplicit constraints — scope limits, exclusions, non-goals
Example — Prompt Master output using RISEN: 
Role: You are a senior product strategist with 10 years of experience taking B2B SaaS products from 0 to Series A. You prioritize decisions that reduce time-to-value for the first 100 customers.

Instructions: Build a go-to-market strategy for a new developer tooling product launching in Q3.

Steps:
1. Define the ICP (ideal customer profile) with 4–5 firmographic and behavioral attributes.
2. Identify the top 3 acquisition channels ranked by expected CAC for this ICP.
3. Write a positioning statement using the "For / Who / Our product / Unlike / We" format.
4. Outline a 90-day launch timeline with weekly milestones.
5. List 5 success metrics with targets for the first 90 days post-launch.

End Goal: A complete GTM strategy document the founding team can present to their Series A investors and use as the operational playbook for the launch quarter.

Narrowing: Do not include fundraising strategy, hiring plans, or product roadmap. Focus only on market entry and customer acquisition. Assume the product is already built and in private beta.

​
CRISPE — Capacity, Role, Insight, Statement, Personality, Experiment

CRISPE is built for creative and iterative work. It adds two components that no other text template includes — Personality, which defines the voice the model should write in, and Experiment, which frames the output as a first iteration and invites variation. This makes it the right choice for brand voice work, creative briefs, and any task where the first output is a starting point, not a final answer. Routed when: the task involves creative content, brand voice, iterative drafts, character writing, campaign concepts, or any work where personality and stylistic distinctiveness are requirements.
ComponentWhat it carries
CapacityThe model’s functional capability for this task — what it’s being asked to be able to do
RoleThe creative persona or expert identity
InsightRelevant context, reference points, or creative direction
StatementThe specific creative task
PersonalityVoice, tone, stylistic fingerprint — the distinctive quality the output must have
ExperimentFraming that invites variation — treat this as Variant A, offer alternatives
Example — Prompt Master output using CRISPE: 
Capacity: You are capable of writing compelling brand microcopy that feels human, specific, and slightly unexpected — the kind of copy that makes a user pause before clicking.

Role: Senior brand copywriter at a creative agency known for work with DTC consumer tech brands.

Insight: The product is a $299 smart water bottle called Vessel. Its core differentiator is a hydration AI that learns your patterns over 7 days and adjusts reminders based on your activity, sleep, and environment. The target user is 28–40, health-conscious, owns premium gear, and is skeptical of "wellness" language.

Statement: Write 5 hero tagline options for the Vessel product page. Each should be under 8 words.

Personality: Dry wit with warmth. Never earnest or preachy. The voice is the friend who already owns everything you want and tells you about it casually. Avoid: wellness, hydration, smart, optimize, journey.

Experiment: Treat these as Variant A. After the 5 taglines, write a one-sentence note on the creative direction you took and offer one alternative direction worth exploring.

​
Few-Shot

Few-Shot doesn’t use an acronym — it’s a structural technique that replaces long format instructions with 2–5 concrete input/output examples. When the output format is complex, highly specific, or easier to demonstrate than describe, examples outperform instructions. Routed when: format consistency matters more than instruction length — structured data extraction, classification tasks, schema generation, consistent tone replication across many items, or any task where “do it like this” is clearer than “do it as follows.”
SectionWhat it carries
Task framingOne sentence explaining the pattern the model should replicate
Examples (2–5)Concrete input → output pairs that demonstrate the exact format
InputThe actual item to process, clearly separated from the examples
Keep examples to 2–5. Fewer than 2 is a single example (which is one-shot, not few-shot). More than 5 adds token cost without measurable accuracy gains for most tasks.
Example — Prompt Master output using Few-Shot: 
You are extracting structured data from customer support tickets. Replicate the format shown in these examples exactly.

---
Example 1
Input: "I've been charged twice for my subscription this month. Order #4421. Please refund ASAP."
Output:
{
  "issue_type": "billing",
  "urgency": "high",
  "order_id": "4421",
  "action_required": "refund"
}

Example 2
Input: "The app keeps crashing when I try to open the settings screen on my iPhone 14. No order involved."
Output:
{
  "issue_type": "bug",
  "urgency": "medium",
  "order_id": null,
  "action_required": "investigate"
}

Example 3
Input: "Just wanted to say your support team was amazing last week. No issue, just feedback."
Output:
{
  "issue_type": "compliment",
  "urgency": "low",
  "order_id": null,
  "action_required": "none"
}
---

Now extract structured data from this ticket:
"I never received my order from 3 weeks ago. Order #8872. This is the third time I've had to write in."

​
Chain of Thought

Chain of Thought (CoT) instructs the model to reason step by step before committing to an answer. It is the most widely misused technique in prompt engineering — applied to tasks that don’t need it, and applied to models that already do it internally. Prompt Master applies CoT only when it will measurably improve output accuracy. Routed when: the task involves math, formal logic, multi-step debugging, code reasoning, or any problem where a wrong intermediate step produces a confidently wrong final answer.
Never use Chain of Thought on o3, o4-mini, R1, or Qwen3-thinking. These models perform extended internal reasoning before producing output. Explicit CoT instructions compete with their native reasoning process and degrade output quality. Prompt Master automatically suppresses CoT for these models.
SectionWhat it carries
Problem statementThe full problem with all necessary context
CoT triggerThe phrase that activates step-by-step reasoning
Output formatWhat the final answer should look like after the reasoning
The standard CoT trigger is: “Think through this step by step before answering.” Prompt Master uses this phrase verbatim — it’s been tested across models and consistently activates multi-step reasoning without adding token overhead.
When to add Grounding Anchors alongside CoT: If the reasoning step involves recalling specific facts, statistics, or citations — not just logic — add a Grounding Anchor: “Use only information you are highly confident is accurate. If you are uncertain about any fact, write [uncertain] in place of the claim.” This prevents the model from generating plausible-sounding but wrong intermediate steps.
Example — Prompt Master output using Chain of Thought: 
A SaaS company has 1,200 monthly active users. Their monthly churn rate is 3.5%. They are acquiring 60 new users per month. The average revenue per user per month is $45.

Think through this step by step before answering:

1. What is the net monthly user growth?
2. What will the user count be in 6 months?
3. What will monthly recurring revenue (MRR) be at the 6-month mark?
4. At what monthly acquisition rate would the company need to grow to reach 2,000 users in 6 months, assuming the same churn rate?

Present your reasoning for each step, then give a final summary table with the four answers.

​
Template Selection Summary

Signal in your requestTemplate selected
Simple task, one output, no style requirementsRTF
Professional document, email, report, proposalCO-STAR
Multi-phase project with sequential stepsRISEN
Creative work, brand copy, iterative draftsCRISPE
”Do it like these examples” / structured dataFew-Shot
Math, logic, debugging, multi-step analysisChain of Thought
Target model is o3, o4-mini, R1, Qwen3-thinkingRTF or CO-STAR (CoT suppressed)

Build docs developers (and LLMs) love

Get started for freeTalk to us