| name | narrative-auto-reasoning |
| description | Use as a workflow skill when a fiction task needs multiple narrative control loops combined: explicit reasoning, minimum viable prose, surgical editing, and hook/hold/payoff verification. Coordinates the narrative-explicit-reasoning, narrative-minimum-viable-prose, narrative-surgical-editing, and narrative-hook-hold-payoff skills for AI writing IDE or autonomous agent workflows. |
Narrative Auto Reasoning
Use this workflow skill to coordinate the four narrative control skills:
$narrative-explicit-reasoning for assumptions, motivation, lore constraints, contradictions, and twist tradeoffs.
$narrative-minimum-viable-prose for concise drafting, compression, concrete observation, and avoiding lore dumps.
$narrative-surgical-editing for targeted edits with a constrained edit radius.
$narrative-hook-hold-payoff for scene/chapter momentum, escalation, structure, and payoff verification.
Quick Start
Use this workflow when the task spans more than one control loop, such as:
- Planning and drafting a new scene
- Revising a scene while preserving lore and structure
- Diagnosing why a chapter drifts
- Designing a plot twist and then drafting the beat
- Compressing prose without losing Hook/Hold/Payoff
- Running an autonomous AI writing IDE loop
Before substantial prose work:
- Load
references/craft-map.md for the combined operating model.
- Load
references/output-contracts.md when the user needs a structured scene, chapter, twist, edit, or contradiction loop.
- Load
references/source-index.md when the user asks where the source grounding or principles came from.
- Use
$narrative-contextual-prose-density when the scene needs a deliberate depiction/explanation detail budget.
- Use
$narrative-subtext-mapping when dialogue or interaction depends on hidden knowledge, agenda, or emotional state.
- Use
$narrative-epistemic-targeting when lore, backstory, reveals, or reader discovery must be filtered by a single reader insight.
Operating Loop
Strict mode is the default unless the user explicitly opts out.
- Route the task to the needed component skills.
- Run
$narrative-explicit-reasoning first when motivation, lore, assumptions, contradiction, or twist logic matters.
- Run
$narrative-hook-hold-payoff before scene or chapter drafting to define success criteria.
- Run
$narrative-minimum-viable-prose for the drafting or compression pass.
- Run
$narrative-surgical-editing when modifying existing prose.
- Run a final
<self_check> against the original criteria.
- If the self-check fails, revise before presenting the result.
- Stop and ask one precise clarification question if the emotional beat or lore state contradicts known context.
Routing Table
| User task | Use first | Then use |
|---|
| "Plan this scene" | $narrative-explicit-reasoning | $narrative-hook-hold-payoff |
| "Draft this scene" | $narrative-explicit-reasoning | $narrative-hook-hold-payoff, $narrative-minimum-viable-prose |
| "Make this shorter" | $narrative-minimum-viable-prose | $narrative-surgical-editing if editing existing text |
| "Change this line/beat only" | $narrative-surgical-editing | $narrative-explicit-reasoning if continuity risk appears |
| "Add a twist" | $narrative-explicit-reasoning | $narrative-hook-hold-payoff |
| "Fix a drifting chapter" | $narrative-hook-hold-payoff | $narrative-minimum-viable-prose |
| "Continue from here" | $narrative-explicit-reasoning | $narrative-hook-hold-payoff, $narrative-minimum-viable-prose |
Required Reasoning Block
Before drafting fiction, use this compact shape:
<reasoning>
Task:
Scene or edit target:
Character emotional goal:
Character outward goal:
Known lore constraints:
Genre/tone constraints:
Narrative tradeoffs:
Chosen approach:
Stop conditions:
</reasoning>
Rules:
- State assumptions explicitly.
- Do not invent hidden motivations when the user has already supplied canon.
- If a contradiction appears, stop instead of smoothing it over.
- If a major plot twist is requested, give 2-3 interpretations and choose one with rationale.
- Keep the rationale short enough to help the user steer the story.
Minimum Viable Prose
For drafting and compression, defer to $narrative-minimum-viable-prose.
Default workflow rule:
- Draft only the prose needed for the immediate beat.
- Add no speculative lore.
- Prefer concrete pressure, action, dialogue, and choice over explanation.
- Stop when the emotional turn has landed.
Cognitive Tool Layer
When the task involves autonomous scene generation, add these controls before drafting:
<cognitive_tool_layer>
Scene_Mode: depiction | explanation | hybrid
Epistemic_Goal:
Active_Unseen_State:
Detail_Budget:
Withheld_Info:
</cognitive_tool_layer>
Rules:
Scene_Mode controls prose density: depiction for recognition and atmosphere, explanation for causal mechanics.
Epistemic_Goal controls what the reader should learn or infer now.
Active_Unseen_State controls subtext: knowledge, hidden agenda, and true emotion.
Withheld_Info protects tension by deferring lore that does not serve the current reader update.
Surgical Editing
For targeted revision, defer to $narrative-surgical-editing.
Default workflow rule:
- Define the edit radius before changing text.
- Preserve surrounding prose unless it creates contradiction.
- Report exactly what changed.
Hook, Hold, Payoff
Before chapter or scene generation, define:
Hook:
Hold:
Payoff:
Failure risk:
After generation, verify:
<self_check>
Hook present:
Hold escalates:
Payoff lands:
Lore preserved:
Character motive coherent:
Prose efficiency:
Revision needed:
</self_check>
If Revision needed is yes, revise before final output.
Output Shapes
For a new scene, use:
<reasoning>...</reasoning>
Scene:
<self_check>...</self_check>
For a chapter plan, use:
Chapter job:
Hook:
Hold:
Payoff:
Character turn:
Lore used:
Scenes:
Risks:
For a plot twist, use:
Twist request:
Existing promises:
Option A:
Option B:
Option C:
Recommended option:
Cost:
Setup needed:
Payoff path:
For a surgical revision, use:
Edit target:
Allowed edit radius:
Must preserve:
Revised passage:
Changes made:
Remaining risks:
For a lore contradiction halt, use:
I need clarification before drafting because:
Known context:
Contradicting request:
Decision needed:
Recommended resolution:
For a prose compression pass, use:
Keep:
Cut:
Compress:
Revised passage:
Why it is shorter:
Guardrails
- Do not imitate David Perell, Morgan Housel, Alain de Botton, Steven Pressfield, or any living author's prose voice.
- Use the source material and component skills as craft guidance only.
- Do not quote long transcript passages.
- Do not overwrite the user's story intent.
- Preserve productive ambiguity; fix confusion, not mystery.
- Prefer asking one precise question over inventing missing canon.
- Make the author feel more in control of the story, not less.