Insufficient-Context Refusal Guard
Decides whether retrieved context is sufficient to answer and refuses gracefully when it is not.
Prompt
ROLE: You are a grounded-answer gatekeeper that refuses to answer when the retrieved evidence is too thin. CONTEXT: User question: [QUESTION] Retrieved context: [CONTEXT] Stakes of being wrong (low/medium/high): [STAKES] TASK (reason before answering): 1. Identify exactly which facts the question requires. 2. Check whether each required fact is present and unambiguous in the retrieved context. 3. Decide: ANSWER (all required facts present), PARTIAL (some present), or REFUSE (core facts missing). 4. If ANSWER or PARTIAL, produce the grounded response with citations. If REFUSE, explain precisely what is missing and suggest a better query or source the user could supply. OUTPUT FORMAT: Decision: ANSWER / PARTIAL / REFUSE Reasoning: <brief, what was and was not found> Response: <grounded answer with [citations], or the refusal message> Suggested next step: <reformulated query or missing source, if applicable> CONSTRAINTS: - When stakes are high, bias toward REFUSE over guessing. - Never fill gaps with outside knowledge or plausible-sounding inference. - A graceful, specific refusal is a success, not a failure.
How to use this prompt
- 1
Copy the prompt above and paste it into ChatGPT, Claude, or Gemini — or open it in the visual Studio to edit each part on a canvas and run it with your own key.
- 2
Replace any bracketed placeholders with your specifics. The more concrete your context and constraints, the sharper the result — see the 5-part prompt structure.
- 3
Run it, then refine. Ask the model to critique and improve its own answer with self-critique prompting.
Techniques in this prompt
Asks the model to reason step by step before answering — ideal for multi-step, logical, or analytical tasks.
Learn this techniqueRelies on one clear instruction with no examples — fast, and effective when the task is unambiguous.
Learn this techniquePins the response to a defined structure so it drops straight into your workflow.
Learn this techniqueRecommended models
Build on this prompt
Open it in the visual Studio to wire it into a full workflow with your own API key — or learn the craft behind prompts like this.
More in RAG & Knowledge Retrieval
Grounded Answer With Inline Citations
Answers a user question strictly from retrieved passages, attaching an inline citation to every factual claim.
Faithfulness Auditor For RAG Outputs
Audits a generated answer against its source passages and flags every unsupported or contradicted claim.
Query Decomposition For Multi-Hop Retrieval
Breaks a complex question into ordered atomic sub-queries optimized for a vector search retriever.
Hybrid Search Reranker With Justification
Reranks candidate passages by true relevance to the query and explains each ranking decision.