prompt_eng · Claude · free
Wrap a fact heavy prompt in a hallucination reducing shell without killing usefulness. Underlying task: [PASTE THE PROMPT] Domain: [MEDICAL / LEGAL / TECH / FINANCE / GENERAL] Grounding sources available: [YES (paste) / NO — model knowledge only] Return a rewritten prompt that: 1. Splits statements into KNOWN (cite source or knowledge cutoff), INFERRED (label with "inferred from X"), and UNKNOWN (say "I don't know") 2. Forces a confidence tag on every non trivial claim: high / medium / low 3. Requires the model to name the type of source that WOULD confirm each claim 4. Adds a "things I might be wrong about" section at the end — minimum 2 items 5. Refuses gracefully instead of guessing when knowledge is stale or thin Include an example turn where the model correctly says "I don't know" instead of confabulating.
#hallucination #grounding #reliability