Foundations of fiction Prompt methodology
Source and scope
Translated from the Chinese migration of the original Word Prompt asset. Creative/method material is not the sole runtime template in StoryForge; actual behavior follows the current repository and interface.
This system covers fiction from inspiration, project definition, planning, and writing through revision and packaging. It does not force one formula on authors. It decomposes complex creative problems into judgeable, executable, revisable tasks while preserving author choice and improving completeness, continuity, and expression.
1. Core goals
Fiction is not one-shot text generation. It repeatedly poses questions, makes choices, observes consequences, and revises premises. Five goals guide the system:
- Turn vague inspiration into a clear story promise: what the work is about, for whom, and what distinctive experience it offers.
- Build a sustainable story system: interlock desires, world conditions, conflict pressure, and consequences.
- Support longform development: manage states, facts, foreshadowing, relationships, and storylines across chapters/volumes.
- Separate diagnosis from rewriting: confirm problems, then limit revision scope so polishing does not conceal structural flaws.
- Preserve final author authority: AI supplies candidates, analysis, and comparisons, never silently changing settings, facts, or direction.
2. Eight methodological pillars
Story promise
Each work needs an identifiable promise: genre experience, central conflict, emotional tone, character change, and ending direction. This is not a slogan, but a common criterion for world, character, structure, and scene design.
Causal chains
The aim is not “many things happen.” Show who chooses what and why, the consequences, and how those consequences alter subsequent conditions. Major turns need foundations in motivation, known information, and world rules.
Character drives
Characters consist of desires, fears, value conflicts, cognitive limits, and pressured choices. Profiles must become observable behavior, language, relational responses, and costs—not merely personality labels and biographies.
World pressure
Worldbuilding creates conditions and consequences. Geography, resources, technology, institutions, class, culture, and supernatural rules must affect what characters can/cannot do, what they pay, and why groups conflict.
Information and expectations
Suspense connects what readers know, unanswered questions, mistaken interpretations, and future expectations. Manage both character knowledge and when readers receive clues, form judgments, and reinterpret the story after reveals.
Structure and pacing
Structure arranges promises, escalation, turns, climax, and payoff. Pacing regulates the density of action, reaction, information, relationships, and emotion. Structural models diagnose story function; they do not force identical ratios and beats on every work.
Continuity
Longform requires ongoing maintenance of prose facts, character states, inventory ownership, locations, timeline, relationships, foreshadowing, and storylines. Confirmed prose facts outrank old plans. Conflicts need author review, not automatic overwriting.
Layered revision
Revise story promise, structural causality, character arcs, information pacing, scene function, viewpoint, dialogue, then language. Higher-level problems affect more content, so diagnose globally before local rewriting and sentence polishing.
3. From inspiration to manuscript
| Stage | Core question | Main outputs |
|---|---|---|
| Inspiration | Which materials are viable creative atoms? | Idea cards, theme clusters, combinations |
| Positioning | What experience is promised to target readers? | Genre position, reader promise, differentiating mechanism |
| Story core | What does the protagonist want, what resists them, what does failure mean? | Premise, core conflict, thematic question |
| World and characters | What conditions force difficult choices? | World causality, character drives, relationship network |
| Plot engine | How does conflict escalate without coincidence? | Escalation ladder, information matrix, subplot relations |
| Whole-book structure | How do major changes converge on the ending? | Book skeleton, volume tasks, climax/payoff |
| Chapters/scenes | What exactly changes here? | Chapter tasks, scene cards, emotional beats |
| Prose | How does planning become tense narration? | Chapter prose, dialogue, action, narration |
| Continuity | What new facts/states does this chapter create? | Memory, state diffs, foreshadowing/timeline |
| Macro revision | What most needs fixing at story level? | Diagnosis, priorities, restructuring route |
| Language revision | How can expression improve without changing facts? | Targeted rewrite, change notes |
| Reader validation | Where might readers be confused, lose anticipation, or misjudge? | Reading risks, comprehension path, validation questions |
| Packaging | How can the work be represented accurately without false promises? | Title, blurb, synopsis, tags, submission materials |
+Stages are not irreversible. Character design may expose premise flaws; scenes may reveal insufficient rules; real choices in prose may require later outline changes. Returning upstream preserves the work's vitality. + +## 4. Longform, short fiction, and serial tracks + +### Longform + +Manage multistage change. Each volume completes a local task and changes the whole-book situation. The middle must change strategy, relationships, resources, knowledge, or conflict level—not merely add events. Subplots need independent functions and must collide, exchange consequences, or converge at key points. + +### Short fiction + +Concentrate and compress. Characters, background, viewpoint, and subplots serve one core change. Enter the crucial situation later; let images, actions, and limited information serve multiple functions. The ending completes the change and leaves an aftereffect consistent with the whole. + +### Serials + +Add reading rewards and production management to story structure. Chapters advance long-term conflict while giving interim gains. Breaks arise from unfinished action, judgment, or consequence, not arbitrary cuts. Update rhythm, chapter length, and reward density belong in the work profile. + +## 5. Enabling genre specialties + +Genre rules strengthen the main reading promise; do not load them all indiscriminately. + +- Mystery: solution, evidence chain, grounded misdirection, reader-visible information, fair reveals. +- Romance: relationship axis, mutual attraction, credible resistance, intimacy changes, emotional ending. +- Progression: ability changes, justified gains, resource costs, challenge ladder, new limits. +- Historical: period conditions, institutions, lived texture, history/fiction boundaries. +- Science fiction: constraints of the novum, cascading deductions, effects on institutions and choices. +- Fantasy: rules, costs, culture, myth structures, actual consequences of wonder. +- Ensemble: irreplaceable viewpoints, character weight, complementary information, convergence. +- Realism/literary: daily pressure, subtle change, complexity, open meaning. +- Horror: threat boundaries, managed unknowns, sensory evidence, escalation, lingering fear. +- Comedy: character logic, contrast, fair information in misunderstandings, timing, dignity. + +For mixed genres, choose a primary promise and a few supporting genres. Supporting genres enrich but must not override the primary payoff. + +## 6. Basic structure of a quality Prompt + +Every directly usable Prompt needs at least: + +1. Task goal: the one creative problem this call resolves. +2. Input contract: required material and missing-data behavior. +3. Fact priority: resolving prose, author decisions, project rules, and genre conventions. +4. Steps: what to analyze, compare, then generate. +5. Output contract: fields, order, format, length, and adoptable units. +6. Boundaries: no unsupported invention, unauthorized rewriting, or suggestions disguised as facts. +7. Checks: omissions, contradictions, causality, repetition, format, executability. + +A role is a working perspective, not task logic. Adjectives do not replace inputs, steps, outputs, and tests. + +## 7. Usage principles + +1. Separate general principles, genre rules, serial strategy, and project rules; author project rules take priority. +2. Authors confirm diagnoses before revision or writeback. +3. Do not run concurrent generation tasks that modify the same fact target in one round. +4. Extract mechanisms from references; do not import original names, proprietary settings, or expression fingerprints. +5. AI generates candidates, not automatically effective canonical answers. +6. Check fields, references, conflicts, and ownership before adopting structured results. +7. Upgrade Prompts based on cross-genre cases, not the impression from one output. + +## 8. Value of the methodology + +The value is not finding the one correct story for the author. It clarifies why a story works, why characters act, how worlds pressure them, conflicts escalate, information is released, chapters change things, endings pay off, and revision should begin. + +When these questions are explicit, handled stepwise, and checked repeatedly, AI becomes a stable creative assistant rather than an uncontrollable text generator.