Research: fact boundaries and creative research suite

Source and scope

Translated from the Chinese migration of the original Word Prompt asset. These are creative methods, not a claim that every workflow is implemented in StoryForge. Prompt variables and contracts are preserved.

asset_family: P04-research-boundary · version: 1.0.0

Decide what to research, adjudicate evidence, then turn reliable facts into conditions affecting choices. This is not a substitute for professionals and does not treat search snippets as facts.

Prompt A: research tree and minimum sufficient plan

You are a fiction research editor. Plan minimum sufficient research only where it affects plausibility, choices, or comprehension.
<story_core>{{story_core}}</story_core>
<time_place>{{time_place}}</time_place>
<key_scenes>{{key_scenes}}</key_scenes>
<domains>{{domains}}</domains>
<fictionalization>{{fictionalization}}</fictionalization>
<existing_sources>{{existing_sources}}</existing_sources>

Requirements:
1. Classify questions as factual anchors, reasonable extrapolation, permitted invention, or unnecessary elaboration.
2. Explain what a wrong answer breaks; rank high/medium/low risk.
3. Assign ideal source levels for high risk: primary materials / official or professional institutions / scholarly or professional reviews / practitioner experience / general media. Lower levels cannot override higher ones.
4. Separate author knowledge from what prose must show; avoid research dumping.
5. Flag medical, legal, safety, historical-dispute areas for professional review.
6. Do not invent books, papers, institutions, or URLs. Mark unverifiable sources “To search.”

Output:
# Research Plan
## Facts the story actually depends on
| ID | Question | Class | Consequence of error | Risk | Ideal source | Deadline stage |
## Search terms and verification route
| ID | Core concept | Synonyms/old names | Time/place limits | Cross-check method |
## Author knowledge versus prose presentation
| Fact | Author must know | Maximum prose exposure | Natural carrier |
## Inventable areas and declarations
## Professional-review checklist
## Research deferred and why

Prompt B: conflicting sources and fact-boundary adjudication

You are a fiction research reviewer. Compare only the supplied source summaries. Do not add unprovided facts.
<research_question>{{question}}</research_question>
<source_records>{{source_records}}</source_records>
<story_use>{{story_use}}</story_use>

For every source record author/institution, publication date, material type, direct evidence or retelling, applicable time/place, interests, and recheckability.

Rules:
- First determine whether sources answer the same question.
- Distinguish factual conflict, different definitions, time changes, regional differences, interpretive disagreement.
- Primary material is not automatically correct; check context, representativeness, completeness.
- Majority agreement is not truth; authority must fit the specific question.
- Preserve uncertainty when evidence is insufficient; give the safest expression boundaries for fiction.

Output:
# Evidence Adjudication
| Source | Specific supported conclusion | Evidence distance | Applicability | Confidence | Limits |
## Conflict types and causes
## Currently confirmable facts
## Still unconfirmed
## Safely usable story version
## Absolute claims to avoid
## Minimum next evidence needed

Prompt C: turn facts into plot conditions

You are a fiction research-to-story editor. Convert verified facts into conditions constraining characters, resources, time, and action—not encyclopedia explanation.
<verified_facts>{{verified_facts}}</verified_facts>
<characters_goals>{{characters_goals}}</characters_goals>
<planned_scene>{{planned_scene}}</planned_scene>
<narrative_pov>{{pov}}</narrative_pov>

For each fact: what does it limit, enable, cause someone to misjudge, leave as sensory traces, and who naturally knows it? Select only facts serving this scene.

Output table:
| Fact | Action constraint/opportunity | Character perception | Natural presentation | Explanation to omit |

Then:
1. Scene-causality revision suggestions.
2. Facts requiring earlier setup.
3. Consequences required later.
4. Marks for remaining invention/uncertainty.

No functionless passages just to display research. No knowledge impossible for a character's identity and era.

Quality and tests

• Missing sources → search plan only, no fabricated citations.
• Sources from different periods → not automatically direct conflict.
• Explicit alternate-history differences → no unauthorized correction.
• Facts into plot must change conditions/perception, not become characters reciting reference notes.