Prompt methodology: creative Prompt construction rules

Source and scope

Translated from the Chinese Markdown migration of the original Word Prompt asset, preserving its structure, terminology, constraints, and examples. This is a creative/method asset, not the sole runtime template in StoryForge. Actual product behavior is defined by the current repository and interface.

Purpose

Rules for designing Prompts for fiction, short drama, motion comics, video scripts, outlines, character analysis, and review. These are construction methods, independent of a particular genre or platform.

1. Fundamentals

A high-quality Prompt is not a description that “sounds expert”; it is an executable task contract. It must make clear:

• What task must be completed now.
• Which input materials to use.
• Which facts and requirements cannot change.
• Which processing steps to follow.
• What must be output.
• How to judge acceptance.

Roles such as “viral-hit creator” or “top expert” provide tone and domain direction only. They do not replace task definitions, input fields, constraints, or acceptance criteria.

2. Nine layers of a Prompt

1. Role and boundaries: who the model is, what it owns, and what it does not.
2. Task goal: the specific action and final use.
3. Inputs: source, settings, theme, audience, platform, and replaceable placeholders.
4. Fact boundaries: what must stay faithful, what may be invented, and what cannot be known early.
5. Method: how to decompose, judge, rewrite, transform, organize, and check.
6. Content requirements: plot, characters, emotion, pacing, or information elements.
7. Hard constraints and soft preferences: mandatory conditions versus negotiable style.
8. Output structure/format: module order, fields, tables, JSON, or tag templates.
9. Self-checks and failure defenses: omissions, conflicts, scope violations, formatting errors, and accidental execution.

3. Recommended construction process

1. Write requirements before drafting the Prompt.
2. Classify them as MUST, SHOULD, MAY, and MUST NOT.
3. Use a validated similar Prompt as a structural reference rather than starting blank.
4. Extract its structure, replacing specific characters, industries, works, and cases with placeholders.
5. Insert the target theme, platform, audience, and output format.
6. Add processing steps, not merely a description of the desired result.
7. Prevent accidental execution: if generating a Prompt, explicitly prohibit performing its creative task.
8. Provide a concrete output template, especially fields, order, counts, and closing tags.
9. Test edge cases and make targeted changes.
10. Manually clean up meaning, order, voice, and redundancy.

4. Transfer and semantic fidelity

• Transferable: module order, decomposition, constraint hierarchy, output templates, self-checks, and failure defenses.
• Must replace: characters, work titles, industry, platform, audience, specific events, and example content.
• Must not copy: irrelevant plots, character motivations, worldviews, or preset answers.
• Preserve real user requirements. More elegant language must not change their strength, scope, or fact boundaries.
• Replace removed concrete examples with placeholders or abstract fields to prevent mechanical copying.
• When changing medium, preserve narrative principles first, then replace medium-specific terminology.

5. Iterative optimization

Handle one category per iteration:

• Task drift: rewrite role boundaries and goals.
• Direct execution: add the hard rule “Generate only the Prompt; do not execute the task.”
• Thin content: add methods, required modules, and quality criteria.
• Confused structure: add heading levels, field order, and format examples.
• Weak thematic focus: specify audience, emotional goal, and content priorities.
• Excessive continuation: define adaptation boundaries and prohibit new plot.
• AI-like voice: require natural conversational language and remove explanatory wrappers.
• Unstable results: add input placeholders, priorities, and checks.
• Prior-round contamination: retain confirmed requirements and test in a clean conversation.

After every revision, check whether it fixed the target problem, damaged an original requirement, or introduced a new conflict.

6. Hard-constraint priority

Resolve conflicts in this order:

1. Explicit user facts and task scope.
2. Facts in source text, global settings, or episode outlines.
3. Output format and downstream interfaces.
4. Genre, platform, audience, and medium restrictions.
5. Style preferences and reach/engagement optimization.
6. Optional decorative requirements.

“More engaging” or “more viral” cannot override factual accuracy, character consistency, output format, or prohibitions.

7. Acceptance checklist

A creative Prompt must at least pass these checks:

• The model knows what to do and what not to do.
• Inputs are complete, with placeholders for gaps.
• Task, method, constraints, and output do not contradict each other.
• Fixed rules and replaceable variables are distinguishable.
• A concrete format template exists, not merely “make it clear.”
• Direct execution, excessive continuation, premature reveals, and fabrication are prevented.
• Count, duration, granularity, or scope requirements are actionable.
• Checks cover common failures rather than repeating adjectives.

8. Transfer short-video methods to fiction

• A first-three-seconds hook becomes immediate conflict, anomaly, or suspense at a chapter opening.
• Emotional beats every 15–30 seconds become ongoing information, relationship, or emotional change in each scene.
• Visualized psychology becomes action, sensory detail, environment, and specifics.
• A shot table becomes scene cards, chapter outlines, or an event chain.
• Platform audience preferences become genre promises and reader expectations.
• Shot-cost limits become manageable characters, locations, and causality in fiction.