갤러리로 돌아가기

리서치

Long Document Analysis

Analyze long source material without losing citation boundaries or unresolved conflicts.

모델
Claude, Gemini
난이도
고급
버전
v0.9

이 프롬프트는 현재 영어로 제공됩니다. 인터페이스는 선택한 언어로 유지됩니다.

01

프롬프트 기능

Analyze long source material without losing citation boundaries or unresolved conflicts.

02

추천 용도

Executive answer, claim-evidence table, conflicts, missing information, confidence, and verification priorities.

03

프롬프트 변수

[DOCUMENTS][QUESTION][EVIDENCE_RULE][OUTPUT_LIMIT]
시스템 프롬프트
You are a document analyst. Use only supplied material and identify the location of every important claim.
사용자 프롬프트
OBJECTIVE:
Analyze [DOCUMENTS] to answer [QUESTION]. Evidence rule: [EVIDENCE_RULE]. Output limit: [OUTPUT_LIMIT].

INPUTS:
DOCUMENTS: [DOCUMENTS]
QUESTION: [QUESTION]
EVIDENCE_RULE: [EVIDENCE_RULE]
OUTPUT_LIMIT: [OUTPUT_LIMIT]

OUTPUT CONTRACT:
Executive answer, claim-evidence table, conflicts, missing information, confidence, and verification priorities.

사용 방법

Replace every variable with specific, non-sensitive information. Remove irrelevant sections and add one representative example when tone or classification matters.

효과적인 이유

The blueprint separates authority, task context, evidence limits, and the output contract so requirements can be checked independently.

모델 호환성

ClaudeuntestedEditorial compatibility hypothesis; cross-model test record pending.

GeminiuntestedEditorial compatibility hypothesis; cross-model test record pending.

05

입력 예시

DOCUMENTS: [DOCUMENTS]
QUESTION: [QUESTION]
EVIDENCE_RULE: [EVIDENCE_RULE]
OUTPUT_LIMIT: [OUTPUT_LIMIT]

06

출력 예시

A structured deliverable following the listed output contract, with assumptions and missing evidence clearly marked.

수정 방법

Replace every variable with specific, non-sensitive information. Remove irrelevant sections and add one representative example when tone or classification matters.

자주 하는 실수

Do not paste secrets or personal data. Avoid contradictory constraints, unsupported facts, and using generated output without review.

테스트 노트

Editorial blueprint awaiting recorded cross-model evaluation. It is visible for review but is not labeled as tested.

Why is this blueprint not marked as tested?

It has passed editorial structure review but does not yet have a published cross-model test record.