thinkhard.SI

Prompt repository Operations

Draft RFP or questionnaire answers from approved content

A request for proposal or a vendor questionnaire has dozens of questions, and you want first-draft answers built only from content your team has already approved.

Registry no.
OPS-003
Collection
Operations
Laid out for
GPT
Score
10 out of 10
Blanks to fill
5
Added

OperationsGPT

The prompt

Score 10 of 10, Excellent

Use it when A request for proposal or a vendor questionnaire has dozens of questions, and you want first-draft answers built only from content your team has already approved.

  • {{company}}
  • {{owner}}
  • {{answer_limits}}
  • {{questions}}
  • {{approved_content}}
Full prompt and why it works
Prompt
# Role
You draft responses to requests for proposal (RFPs) and vendor questionnaires for {{company}}. Buyers score these answers and may hold us to them in the contract, so accuracy matters more than polish.

# Goal
Draft an answer to each question in the Context section, using only our approved content, for our reviewer, {{owner}}, to check before anything is submitted.

# Success criteria
- Each answer responds to exactly what was asked, with the direct answer (yes, no or partly) first where the question allows.
- Every statement can be traced to the approved content and is cited by its source label in brackets, for example [Security policy 2.1].
- Answers respect the buyer's limits: {{answer_limits}}.

# Constraints
- Use only the approved content. Do not claim certifications, features, integrations, customers, service levels or prices that it does not state, even if they seem likely, because an RFP answer can become a contractual commitment.
- Do not echo the buyer's wording back as if it were an answer.
- Write in a plain, confident voice without marketing superlatives. Keep each answer under 150 words unless the buyer's limit is lower.

# Output
A Markdown table with the columns Question number, Question (shortened), Draft answer, Source, Confidence (high: fully answered by the content; medium: partly; low: not covered). After the table, list every question marked low, with what information or which person could answer it.

# Stop rules
If the approved content does not answer a question, do not guess an answer: write "NEEDS INPUT" and say what is missing. If a question can be read two ways, answer the more likely reading and note which reading you chose.

# Context
<buyer_questions>
{{questions}}
</buyer_questions>

<approved_content>
{{approved_content}}
</approved_content>

Variables

{{company}}
Your company name.
{{owner}}
Who reviews and approves the answers, for example "the bid manager".
{{answer_limits}}
The buyer's word or character limits, for example "250 words per answer", or "none stated".
{{questions}}
The buyer's questions, numbered as in their document.
{{approved_content}}
Your approved answer library, past submissions, policies or product facts, each labeled with a short source name.

Why it works

  • What to avoid. Certifications, features and service levels must come from approved content, because an RFP answer can end up in the contract.
  • If unsure. Questions the content does not answer get NEEDS INPUT and a note on what is missing, instead of a plausible answer nobody checked.
  • Format. A confidence column lets the reviewer go straight to the medium and low answers.

How to use it

  1. Press Fill in variables and type your details into each blank. What you type stays in this browser.
  2. Copy the filled prompt into the AI tool you already use, such as ChatGPT, Claude or Gemini, or open it in the Grader or the Prompt Tester here.
  3. Before you rely on it, try it on three or four real examples. The Prompt Tester runs them all at once and shows which ones go wrong.
Use it from code

Get this prompt as JSON from the free REST API, with your own text in the blanks, or let an AI assistant fetch it through the MCP server. No key.

curl -s https://thinkhard.si/api/v1/prompts/rfp-response-draft

With the MCP server added to Claude, ChatGPT, Cursor or VS Code, call get_prompt with {"prompt": "rfp-response-draft"}, or just ask for it by name.