Master blueprint: fiction stages and the Prompt system

Source and scope

Translated from the original Word asset's Chinese Markdown migration. This is a method asset, not StoryForge's sole runtime template. Product coverage and gap assessments below describe the source document's context; consult the current feature guides for present capabilities.

Version: 1.0

Scope: longform fiction, short fiction, serialized web fiction.

Purpose: define asset classes, dependencies, inputs/outputs, and construction order—not a fixed writing formula.

1. Design conclusions

1. Fiction cycles among conception, planning, drafting, checking, and revision. Stage numbers express dependencies, not a mandatory linear sequence.
2. Longform and short fiction are not length settings on one template. Longform needs volume promises, woven subplots, long-term escalation, and continuity ledgers. Short fiction needs one core change, limited cast/viewpoints, compressed background, and earlier irreversible change.
3. Serial strategies are not universal laws. End hooks, fixed reward frequency, paywall breaks, and steady updates apply only with a serial profile; do not impose them on literary shorts, non-genre fiction, or complete manuscripts.
4. Structure models diagnose and plan; they are not fill-in formulas. Select three-act, seven-point, story-circle, snowflake, scene/sequel, etc. for the project's problem, not all at once.
5. Prompts serve author decisions. Quality output offers comparable alternatives, explicit tradeoffs, and adoptable structured results—not ornate terminology instead of concrete plot.

2. Four rule layers

Layer	Meaning	Examples	Default
U Universal narrative principles	Broadly useful, but intentional violation allowed	Understandable causality, motivated behavior, setup/consequences	On
G Genre conventions	Meet particular reader expectations	Fair mystery clues, romance relationship axis, perceptible progression	On for selected genre
S Serial strategy	Chapter reading, following updates, platform rhythm	Chapter reward, interim promises, update buffer, breaks	Serial mode only
P Project preferences	Author-chosen style/limits	Viewpoint, exclusions, density, tone, ending direction	Highest priority

Conflict order: facts/written prose > explicit author requirements > project rules > genre conventions > serial strategies > universal defaults. “Universal” remains a default suggestion and cannot override deliberate experimentation.

3. Overall flow

Materials → clarify inspiration → positioning/promises → story core → world/research → characters → plot engine → structure → longform/short branch → chapters/scenes → prose → continuity → macro revision → scene/language revision → reader validation → packaging.

Two cross-cutting tracks:

• Facts/continuity: character states, timeline, inventory, locations, world rules, foreshadowing, written facts.
• Quality assurance: copyright/compliance, information security, versions, test cases, failure repair.

4. Stage task matrix

P00 Creative task definition

• Goal: turn “I want to write something” into an executable brief.
• Required: intent, length mode, readers, serial status, untouchable boundaries.
• Output: brief, known/unknown list, decision priorities, risks.
• Prompt families: requirements clarification, constraint organizer, missing-information diagnosis.
• Source-document product mapping: project information, creative rules, references.

P01 Raw inspiration and material cleanup

• Goal: retain transferable conflict, imagery, relationships, mechanisms, and emotion without source fingerprints.
• Required: fragments, notes, dreams, observations, references.
• Output: atomic idea cards, mechanisms, variables, non-reusable elements, risks.
• Families: atomization, decontextualized abstraction, cross-material clustering, contradiction merging, follow-up questions.
• Mapping: inspiration, references, creative rules; the source assessment emphasizes recording and identifies clustering/selection gaps.

P02 Genre positioning and reader promises

• Goal: identify which readers keep reading, why, and the experience delivered.
• Required: idea cards, length mode, candidate genres.
• Output: genre/subgenre, readers, core gratification/emotional value, promises, differentiation, banned tropes.
• Families: genre identification, expectation matrix, promise/payoff table, differentiation, comparable-work risk checks.
• Caution: market positioning is a choice, not permission to flatten author originality.

P03 Concept selection and story core

• Goal: condense ideas into concepts with causal potential.
• Required: idea cards, genre promises, author preferences.
• Output: logline, concept, protagonist, desire, resistance, cost, thematic tension, dramatic question.
• Families: divergence, scoring, premise stress test, core generation, thematic opposites.
• Mapping: theme, conflict, concept, main plot/subplots in story core.

P04 Research and fact boundaries

• Goal: distinguish required research, reasonable extrapolation, and permitted invention.
• Required: story core, time/place, professions/technology/culture.
• Output: research tree, fact anchors, source levels, verification gaps, alternate-history declarations, anachronism checklist.
• Families: research planning, historical verification, professional-scene consultation, conflicting-source adjudication, facts into plot.
• Mapping: historical materials, world rules, locations, encyclopedia entries, scene research.

P05 World and narrative environment

• Goal: build what affects choices, resource distribution, and conflict, not an encyclopedia pile.
• Required: core, genre, research boundaries, initial character needs.
• Output: rules, spatial relations, resources/livelihood, institutions, factions, historical causality, customs, sensory life, conflict sources.
• Families: principles, geography-resource-society chains, institutional pressure, faction ecology, timeline, lived texture, contradiction audit.
• Mapping: origins, nature/culture, power systems, geography, history, rules, locations, entries.

P06 Character system

• Goal: a network making distinctive choices under pressure.
• Required: core, world pressure, genre promises.
• Output: role, external goal, internal need, values, fears, contradictions, resources, weaknesses, voice, arc, relationship dynamics.
• Families: protagonist/opposition/supporting design, contrasts, lie-desire-need, relationship tension, voice, redundancy audit.
• Mapping: character fields, relationships, states, character-driven plotting.

P07 Plot engine

• Goal: continuously produce choices, resistance, consequences, and new problems.
• Required: goals/resources, world rules, genre promises.
• Output: conflict chain, escalation ladder, risk/cost, opposition strategies, information asymmetry, failure consequences, repeatable variable mechanisms.
• Families: goal-motivation-conflict, antagonism, escalation, try-fail cycles, information release, rewards/costs.
• Mapping: core, storylines, emotional beats, power systems; source assessment notes gaps in dedicated escalation and information matrices.

P08 Structure selection and whole-book turns

• Goal: select the smallest sufficient structural toolkit.
• Required: core, intended ending, length mode, genre promises.
• Output: selection rationale, opening state, inciting event, turns, midpoint change, crisis, climax, new ending state.
• Families: model recommendation, backward ending design, major beats, comparative audit, anti-formula rewriting.
• Constraint: explain fit and non-fit; never force multiple templates simultaneously.

P09-L Longform architecture

• Goal: expand major turns into sustainable volumes, storylines, and promise/payoff systems.
• Required: turns, target length, serial mode, main arcs.
• Output: volume goals/conflicts, end-state changes, subplot weaving, climax gradient, middle support, foreshadowing cycles, information plan.
• Families: volume structure, middle diagnosis, weaving, arc alignment, climax ladder, volume promise ledger.
• Mapping: volume/chapter plans, storylines, foreshadowing, emotional beats; source assessment calls for stronger subplot collisions and volume ledgers.

P09-S Short-fiction architecture

• Goal: compressed narrative around one core change.
• Required: core event/epiphany, protagonist, dominant emotion/insight, target length.
• Output: minimal cast, single plot, entry point, key choice, turn, ending aftershock, background to delete.
• Families: single effect, compression, enter late/leave early, character merging, reflective endings, deletion audit.
• Constraint: no subplots or vast settings by default; add a second viewpoint/frame only when necessary.

P09-R Serial production

• Goal: chapter-reading experience and production buffers without damaging integrity.
• Required: longform architecture, update frequency, chapter-length range, platform/reader habits.
• Output: interim rewards, update buffer, volume/chapter breaks, follow-up questions, feedback-adoption rules.
• Families: serial rhythm, end breaks, promise density, draft buffer, feedback denoising.
• Constraint: end hooks/update rhythm become hard constraints only in serial mode.

P10 Chapters and scenes

• Goal: irreplaceable narrative function for every chapter/scene.
• Required: storyline, adjacent beats, current states, scene conditions.
• Output: chapter function, scene goal, resistance, turn, result, emotional change, new information, entry/exit states, estimated length.
• Families: chapter task sheet, scene cards, scene/sequel, emotion beats, information increments, delete/merge diagnosis.
• Mapping: chapter plans, beat outlines, scenes, emotion arcs, opening continuity, endings.

P11 Prose drafting

• Goal: concrete action, senses, voice, and rhythm from confirmed plans.
• Required: chapter task, scene cards, continuity facts, voices, creative rules.
• Output: prose draft and uncertainty list if needed.
• Families: first chapter, regular chapter, continuation, dialogue, action, interiority, description, transition, chapter titles, alternate opening/ending.
• Constraint: choose showing, summary, or omission by narrative purpose rather than absolute “show, don't tell”; no default suspense ending every chapter.

P12 Continuity and narrative memory

• Goal: consistent facts, states, inventory, time, locations, relationships, foreshadowing across chapters.
• Required: prose, previous handoff, plan, fact ledger.
• Output: summary, ending handoff, plan/prose deviations, state changes, timeline events, item transfers, foreshadowing changes, conflict alerts.
• Families: chapter memory, fact extraction, states, timeline checks, inventory, relationships, foreshadowing lifecycle, plan reconciliation.
• Mapping: identified in the source assessment as the strongest existing area and a mandatory input layer for longform Prompts.

P13 Macro revision

• Goal: fix story before sentences.
• Required: manuscript/layered summaries, promise, structure, reader feedback.
• Output: evidence, scope, priority, preservation, delete/change/move/add options, order.
• Families: developmental editing, causality, weak middle, drifting arcs, detached subplots, weak climax, ending payoff, restructuring.
• Constraint: diagnose separately from rewriting; cite locations/symptoms before proposing fixes.

P14 Scene and language revision

• Goal: improve scene function, viewpoint, precision, and voice without unauthorized plot changes.
• Required: text, permitted scope, context, style rules.
• Output: issues, targeted revision, change notes.
• Families: tension, subtext, viewpoint/distance, pacing, description ratio, redundant information, polishing, expansion/compression, de-templating.
• Constraint: ornate is not necessarily better; expansion must add narrative function.

P15 Reader experience and validation

• Goal: simulate comprehension, expectations, emotion, and abandonment risks for different readers.
• Required: text, readers, promises, review dimensions.
• Output: reading path, question timeline, expectations/payoffs, confusion, credibility issues, severity, revisions.
• Families: opening blind read, genre readers, beta simulation, suspense fairness, emotional credibility, readability, excessive trope-subversion diagnosis.
• Constraint: risk signals only, never pretend to be real market data.

P16 Titles, blurbs, and packaging

• Goal: accurately convey promises, not conceal content mismatch.
• Required: core, reader position, completed content, spoiler limits.
• Output: book/chapter titles, blurbs, synopses, tags, submission summaries in different formats.
• Families: title divergence/selection, chapter titles, spoiler-free blurb, complete synopsis, selling points, packaging consistency.

P17 Archive and reuse

• Goal: turn effective decisions and runs into iterative assets.
• Required: adopted outputs, human revisions, failures, scores.
• Output: versions, conditions, counterexamples, evaluations, improvement rules.
• Families: run review, difference attribution, Prompt repair, rule upgrades, deduplication.

5. Source assessment: coverage and gaps

Strong coverage

• Worlds, characters, story core, creative rules.
• Volume/chapter/beat outlines, scenes, character-driven plots.
• Prose, continuation, polishing, expansion, de-templated expression.
• Memory, facts, states, items, locations, timeline, relationships, foreshadowing.
• Scene research, historical consultation, style learning, multiple reviews.

Priority gaps identified by the source

1. Inspiration cleanup/clustering, concept selection, briefs.
2. Explicit genre-promise, reader, and length-mode contracts.
3. Structure-model selection rather than one rigid web-fiction template.
4. Subplot weaving, information release, climax ladders, volume promises, middle diagnosis.
5. Independent short-fiction architecture.
6. Chapter titles, opening blind reads, ending payoff, developmental editing.
7. Reader simulation, feedback denoising, evaluation sets, version governance.

6. Asset types

Each stage can contain at least one:

• Generation: candidates from confirmed inputs.
• Diagnosis: locate issues without automatic rewriting.
• Revision: targeted change from diagnosis and allowed scope.
• Decision: compare options consistently, preserving author choice.
• Extraction: structured facts from prose/material.
• Validation: facts, format, creative constraints.
• Meta-Prompt: generate/adapt the above from specifications.

Do not create multiple assets merely by changing titles. If inputs, logic, and output fields match, merge into one parameterized Prompt; different creative goals or validation criteria require separation.

7. Initial construction order

Batch 1: decision foundations

P00, P01, P02, P03, P08, and quality standards: ensure later work solves the right problem.

Batch 2: longform core

P06, P07, P09-L, P10, P12, P13: choices, plot engines, sustained volumes, continuity, macro revision.

Batch 3: short-fiction and serial branches

P09-S, P09-R, opening/chapter breaks, short compression, serial feedback denoising.

Batch 4: genre specialties

Mystery clue matrices, romance relationship beats, progression economies, historical research, ensemble/viewpoints. Each declares scope and must not become a universal rule.

Batch 5: language and release

Scene revision, dialogue, narrative distance, chapter titles, blurbs, synopses, beta simulation, asset retrospectives.

8. Completion criteria

All required:

• Clear goal, required/optional inputs, structured output.
• Longform, short, serial, and genre differences distinguished.
• Maps to existing data or explicitly marks a product gap.
• Tests for normal, missing input, conflicting input, and continuity conflict.
• No traceable source-work content, private information, or unauthorized long quotations.
• At least one independent reread; no duplicated core content across assets.