Yanxuan Story Creation System
5 Foundation Files + 9-Step Pipeline · Industrialized Short Suspense-Story Production Line
An industrialized creation system for high-completion-rate suspense / twist short stories on paid-story platforms like Yanxuan. Its core paradigm is "5 Foundation files + 11 gates + 9-step pipeline": it hard-codes front-loaded constraints across the plot_summary → world / characters / outline / voice / canon five phases, then drives the story from setting to a publishable draft through a 9-step pipeline — Foundation alignment → plot generation → initial generation → scene generation → editorial review → reader feedback → author revision → logic verification → proofreading gate → style rewrite. With 11 Foundation gates (F1–F11), a 10-question adversarial test, and multi-author style-fingerprint rewriting, it upgrades "human reviewing human" into "machine gates, human decides."
Related Systems
YanXuan focuses on "short suspense + industrialized pipeline," while novel-writing focuses on "long-form + 30-step ordered flow" — the two form a short / long complementary pair.
Core Modules
Books Overview · 13+ Story Samples with Chapter Stats
Foundation Review Gates · 5 Phases, 11 Gates + 10-Question Adversarial Test
Pipeline Run Control · 9-Step Conveyorized Scheduling
Core Philosophy
Decompose "author inspiration → publishable draft" into 5 Foundation files + 11 gates + 9-step pipeline — front-load constraints, automate review, and make style reproducible.
5 Foundation Files Lock Front-Loaded Constraints
Before writing a single word, you must complete five files — world.md (physical constraints & sensory map) / characters.md (character arc & villain disguise) / outline_v2.md (foreshadowing / event function / chapter BEATS) / voice.md (style fingerprint & Exemplars) / canon.md (timeline & evidence chain). All constraints are locked before drafting, so downstream never "fixes the setting after the fact."
11 Gates F1–F11 · Machine Gatekeeping
F1 physical constraints / F2 character arc / F3 sufficient foreshadowing / F4 increasing event function / F5 style fingerprint / F6 Canon timeline / F7 evidence chain / F8 Exemplars quality / F9 hook effectiveness / F10 villain disguise layers / F11 logic confrontation — each gate has a corresponding file and fix method; failing a gate blocks entry to the next phase.
9-Step Pipeline + Author–Editor–Reader Iteration
Foundation alignment → plot generation → initial generation → scene generation → editorial review → reader feedback → author revision → logic verification → proofreading gate → style rewrite. Author / editor / reader iterate over multiple rounds with different prompts, turning single-shot generation into a traceable engineering flow.
Style Fingerprint + Multi-Author Rewrite
Extract the author's style profile (sentence length / dialogue ratio / sensory allocation / rhythm features) from samples into profiles/, then rewrite the text by fingerprint — the same story can switch to the "Shisanshua" style or a custom target author with one click.
Industry Pain Points
Industrialized production of short suspense / twist stories has long faced six common pain points — each one directly drags down completion rate and publish rate.
Reliance on Personal Inspiration & Feel
Creators start on a "stroke of genius" with no methodology; hits are unrepeatable, team throughput is unpredictable, and newcomers take long to ramp up.
Foreshadowing / Canon Contradictions Found Only After Writing
Traditional flow is "write first, ask later," leading to uncollected foreshadowing, timeline conflicts, and broken evidence chains; fixing one thing breaks everything, and bugs surface only before publishing.
Severe AI Writing Style Drift
AI-generated text carries a heavy "AI smell" — sentence length / dialogue / sensory choices diverge from the target author's style, forcing authors to spend huge effort hand-editing.
No Unified Standard for Multi-Author Collaboration
When a team co-writes, each person records character voice, foreshadowing, and Canon differently; merging causes style breaks and setting conflicts.
Subjective, Non-Reviewable Evaluation Criteria
Assessments of completion rate / hook / twist rely entirely on the editor-in-chief's subjective judgment, lacking a quantifiable, reviewable checklist.
Volatile Platform Tone & Compliance Constraints
Different paid platforms (Yanxuan / Fanqie / Qidian) differ sharply in tone, length, and compliance red lines; when platform rules change, large numbers of old drafts become obsolete.
Core Capabilities · 9 Functions
Designed around the full chain "5 Foundation files → 11-gate review → 9-step pipeline → style rewrite," each capability has clear input / output and responsibility boundaries.
Structured Construction of 5 Foundation Files
world / characters / outline_v2 / voice / canon each have explicit templates and fill-in steps (plot_summary lock → physical constraints → sensory map → alibi → evidence chain → internal consistency); each file can be checked and back-filled independently.
11 Foundation Review Gates
F1–F11 check quantifiably item by item (foreshadowing ≥10, density ≥5 per 1k words, ≥3 indirect-characterization methods, ≥3 villain-disguise layers, etc.); failing a gate forces rework with a clear fix.
10-Question Adversarial Test
"Refute your core premise in one sentence?" "Does the story still hold if the villain is an ordinary person?" "Can the first 100 words make a stranger curious?" "Does the ending take 10–15%?" — ten questions reverse-interrogate the setting to surface fatal flaws early.
9-Step Creation Pipeline
Foundation alignment → plot generation → initial generation → scene generation → editorial review → reader feedback → author revision → logic verification → proofreading gate → style rewrite; each step has an independent prompt, independent log, and independent exit code.
Author–Editor–Reader Iteration Loop
Author drafts → Editor gives structure / rhythm / foreshadowing / voice feedback per the editor prompt → Reader gives "completion rate / hook / immersion" feedback per the reader prompt → Author revises per feedback; multiple iterations turn single-shot generation into engineering polish.
Logic Verification + Canon Timeline Check
An independent logic-verification script (logic_verification_prompt) checks item by item across four dimensions — motive / timeline / evidence / caliber; the Canon timeline is precise to the hour, and the evidence chain is fully passed from creation to current holder.
Style Profiling & Multi-Author Rewrite
Extract the author's style fingerprint from samples (reference/) into profiles/ (templates like shisanshua_style / hard-boiled noir / lyrical plain-sketch / playful dialect), rewrite text by fingerprint, and support line-by-line diff between any two drafts.
Auto-Generated Story Notes
One click generates a publish-ready "story notes" file + chapter titles for unified external publishing; golden_12 / The Third List already have the standard trio — notes + chapter names + creation log.
One-Stop Visual Admin Console
Flask + vanilla JS, no-build admin console with four modules — books overview / Foundation review / style rewrite / pipeline run; paths are locked inside the project to prevent traversal, and the backend reads only without polluting source files.
9-Step Creation Pipeline · From Foundation to Publish
After Foundation alignment, the pipeline runs; the 9 steps execute in dependency order, each independently runnable with independent logs.
Key Results · Foundation & Pipeline Assets
Crystallize the capability of "industrialized short-suspense-story production" into measurable, reusable engineering assets.
Key Breakthroughs
The difficulty is not in single-point rules, but in stringing "Foundation–Review–Pipeline–Style" into one traceable, reviewable, reusable industrialized production line.
Front-Load Constraints, Not After-the-Fact Fixes
The 5 Foundation files lock physical constraints / character arc / foreshadowing / style / Canon before writing, so downstream writing never "fixes the setting after the fact" — eliminating at the source the engineering problems of "uncollected foreshadowing" and "Canon contradictions" that only surface after writing.
11 Gates + 10-Question Adversarial Test
F1–F11 are machine-checkable, quantifiable rules (foreshadowing ≥10, density ≥5 per 1k words, ≥3 indirect-characterization methods, ≥3 villain-disguise layers); the 10-question adversarial test reverse-interrogates the setting; together they upgrade "subjective review" into "automated review."
Author–Editor–Reader Iteration Loop
The author drafts per Foundation; the editor gives structure / rhythm / foreshadowing feedback per the editor prompt; the reader gives completion-rate / hook feedback per the reader prompt; multiple iterations upgrade "AI writes it all at once" into "AI participates in engineering polish."
Reproducible Style Fingerprint
Extract the author's style profile from samples into profiles/, then rewrite text by fingerprint; the same story can switch to "Shisanshua" or a custom target author with one click, solving the industry pain of "heavy AI smell / large hand-editing burden."
Business Value
Convert technical capability into quantifiable, reusable business value.
Content Under Active Update
Product content has been published based on internal materials. The following areas are planned for further development:
Other Products & Solutions
Explore Xianma AI solutions in other domains