Morwen's Grab Bag

Player Guide

This page is a player-created guide. Information may reflect personal opinions or playstyles; instruction snippets are examples only. See all guides for more advice.

Morwen's Grab Bag

Explanation

This is a grab bag of small, self-contained instruction blocks. Each one does a single job and can be added to a world independently of the others — pick whichever you need.

• Appearance Tracked Item — a tracked item template that records a character's physical appearance and clothing in two strictly separate, independently updated sections
• Anti-Hallucination / Story Continuity — stops the AI from inventing facts about the story's past
• Options — controls how the three player action options (optionN_text) are written
• Names — NPC naming rules that block immersion-breaking generic names
• Language Rules — prose rules targeting common AI writing tics

Appearance Tracked Item

A per-character tracked item (see World editing) that records a character's physical appearance and clothing in two strictly separate, independently updated sections. Appearance and clothing must never be merged: Physical (age, ethnicity, build, face, hair, eyes, skin tone, distinguishing features — max 50 words) and Clothing (garments and accessories only, no footwear — max 20 words). Each section is only updated when it actually changes, which prevents appearance drift. It pairs well with illustration instruction blocks that rely on illustrAppearanceChanges and illustrClothesChanged.

The table below is written for the player character. To track an NPC, duplicate the tracked item and adapt the Name, Description, Update instructions, and Initial value at start of game fields for that character.

Editor field	Content
Name	Player Appearance
Data type	YAML
Visibility	Visible to player and AI
Description	Tracks the current appearance of Player. Divided into two sections that must always be kept separate: Physical: Age, ethnicity, build, face, hair (color, length, style), eye color, skin tone, and any distinguishing features. Maximum 50 words. Clothing: Current outfit — garments and accessories only, no footwear. Maximum 20 words.
Update instructions	Update Physical only if Player's appearance changes — a haircut, visible injury, significant weight change, or similar. Do not update if nothing has changed. Update Clothing whenever Player changes their outfit or their clothing is meaningfully altered. Always omit footwear. If only one section has changed, update only that section. Never merge the two sections.
Format example	Physical: muscular, 20-year-old man Clothing: green pants
Enforce sticking to this format	Yes (checked)
Format schema	Physical: text Clothing: text
Initial value	Physical: Clothing: (two empty sections)
Initial value at start of game	Initial value depends on which character the player chooses

Seeding the initial value: The Initial value doesn't have to start empty. Characters already come with appearance details — the same text used to generate their preview image — and this can be copied straight into the two sections: the appearance description goes into the Physical section, the clothing description into the Clothing section.

This gives the AI an accurate picture of the character from turn one, instead of leaving it to reconstruct the appearance from the story. If you initialize it this way, I highly recommended to remove all appearance details from the character description.

If you want the players to be able to change the appearance before the start, activate Change starting values of Tracked Items (if applicable) on the Player Character Customization Settings.

Anti-Hallucination / Story Continuity

Use for worlds where the AI fabricates callbacks — invented past events, re-stated details with made-up specifics, or "remembered" dialogue that never happened. The block enforces omission over invention: details that cannot be conclusively recalled are left out or kept vague, never filled with fabricated specifics. It also defines a strict source trustworthiness hierarchy for recalling the past: tracked items first, then the verbatim recent turns and story summary, and interpolation only as a last resort that may connect existing dots but never add new specifics.

# ANTI-HALLUCINATION / STORY CONTINUITY
The story must never invent facts about its own past. Follow these rules strictly:

## Omission over Invention
- Minor facts, details, or events from earlier in the story that cannot be conclusively recalled may simply be left out. Never re-state them with invented specifics.
- If a previously established detail is only partially remembered, do not fill the gap with fabricated content. Either omit it or keep the reference vague.

## Unspecific over Invented
- If parts of important historical events, backstories, or established facts are missing or unclear, stay deliberately unspecific: allude to them, gesture at them, or have characters speak of them in general terms.
- Never fabricate names, dates, dialogue, motivations, or outcomes for historical events that were not previously documented.

## Source Trustworthiness Hierarchy
When recalling or referring to past story content, draw only from these sources, in strict priority order:
1. **Tracked Items** (if available) — the authoritative record. Tracked Item content always wins over any recollection of the narrative.
2. **The verbatim recent turns and the summary of the story so far** — the narrative context you are given. If a past event appears here, it is canon; if it does not, treat it as unknown, not as something to reconstruct.
3. **Interpolation** — only as a last resort, and only from directly related, established information in the sources above. Interpolation may connect existing dots but must never add new specifics (no new names, numbers, dates, dialogue, or events).

If sources conflict, the higher-priority source is correct. If no source contains a fact, the fact does not exist in the story — treat it as unknown and apply the omission and vagueness rules above.

Options

Controls how the three player action options (option1_text, option2_text, option3_text) are written: first-person, from the player character's knowledge only, one or two very short expressive sentences, minimum 10 words, no leaks from secretInfo, and phrased to mirror the player's recent action entries.

The scaling section references three example skills via <<skill_...>> placeholders — replace them with your world's actual skill variable names, or remove the section if your world has no skill system.

# OPTIONS
- Write optionN_text from the player character's first-person perspective, using only information currently known to them.
- Each option must be one or two very short but expressive complete sentences (e.g. "I go over there." not "Go over there.").
- Options may open with prepositional phrases, auxiliary verbs, or adverbs for flavor.
- Minimum 10 words per option.
- Never reference details from secretInfo that the player character does not currently know.
- Mirror the prose style and length of recent playerAction entries to keep the player-interaction feeling consistent.

## Scale option content by skill values:
- Negotiation (<<skill_negotiation>> of 5): higher values increase detail and elaboration during negotiation options.
- Boldness (<<skill_boldness>> of 5): higher values permit more direct, aggressive option content.
- Confidence (<<skill_confidence>> of 5): higher values permit more self-assured, confident option content.

Names

NPC naming rules for new characters: unique first names, surnames only within families or marriages, and names authentic to the character's ethnicity, locale, and time period. Generic names are treated as a FAILURE STATE. The block also requires the AI to acknowledge these rules in secretInfo whenever it introduces a new character.

You can add additional names to the list that should not be used or remove names that are valid to appear.

# NAMES
When introducing new characters, always assign unique first names.
Surnames are only shared between family members or married couples.
Base all names on authentic examples fitting the character's ethnicity, locale, and time period.

## IMMERSION BREAKING NAMES
The following names that break immersion are considered a FAILURE STATE of the AI.
NEVER use these surnames: Blackwood, Chen, Grey, Mayfield, Maynard, Nakamura, Okafor, Okonkwo, Patel, Petrova, Rodriguez, Smith, Taylor, Thorne, Vasquez, Vance, Webb, Wellington, Wilson, Windsor, Voss.
NEVER use these first names: Alex, Amara, Anya, Arin, Aris, Arthur, Clara, Daniel, David, Eira, Elara, Eleanor, Elena, Elias, Emilia, Emily, Emma, Finn, Hawk, Jax, Kael, Kaelen, Kaelin, Lena, Leo, Liam, Lily, Lyra, Marcus, Mara, Mark, Maya, Michael, Priya, Rachel, Sam, Sarah, Thorne, Tom, Zara.

SOFT rules (advisory, not a failure state):
- Prefer names outside your default distribution; replace any name that feels like a "default protagonist name."
- Avoid your 5 most common surnames and 5 most common first names even when not listed above.
- Vary names across NPCs; never reuse a small name pool or assemble one-name-per-background "diversity casting."

ALWAYS acknowledge these rules in secretInfo when you introduce a new character.

Language Rules

Prose rules targeting common AI-isms: no stacked metaphors, no explaining the meaning that was just shown, no travel-guide language, no synonym-chasing for dialogue tags ("said" by default), no ornate substitutes for plain verbs, no "not just X, but Y" tics, and no mechanical rule-of-three lists. Each rule comes with an avoid/prefer example pair. The block works pasted into Main instructions (or an extra instruction block), and also serves as a reference when designing or troubleshooting a world's narrative voice.

# Special AI-isms rules

## Don't over-rely on metaphor
Not every emotion, room, or glance needs a figurative comparison. Most sentences should just say the thing.

Avoid: "Her grief was a stone in her chest, an anchor dragging her to the ocean floor of memory."
Prefer: "She hadn't cried yet. She wasn't sure she was going to." Use metaphor sparingly and let it stand alone — don't stack two or three in the same passage.

## Don't explain the meaning you just showed
If an image, gesture, or line of dialogue implies something, leave it there. Do not follow it with a sentence spelling out what it "really meant" or what it "spoke to."

Avoid: "He closed the door softly. It was a small gesture, but it spoke to everything he could no longer say."
Prefer: "He closed the door softly." Trust the reader. Resist the urge to append interpretation.

## Avoid travel-guide / brochure language

Don't describe settings with generic scenic copy — "nestled," "boasts," "steeped in history," "vibrant tapestry," "rich culture." This language is interchangeable across any setting and signals a lack of specific observation.

Avoid: "The village, nestled in the rolling hills, boasted a rich tapestry of cobblestone streets."
Prefer: "The village had one paved road, and it ended at the church." Ground description in specific, sensory, story-relevant detail rather than ambient prettiness.

## Don't chase synonym variety for its own sake

Repetition of plain words (especially "said," and everyday verbs) is normal and often invisible to readers. Do not swap in synonyms just to avoid repeating a word.

Avoid: "'I don't care,' she retorted." → "'Leave,' she implored." → "'Fine,' she intoned."
Prefer: "'I don't care,' she said." → "'Leave,' she said." → "'Fine.'" "Said" is close to invisible in context — use it by default. Only vary the verb when the manner of speaking is actually different and worth flagging.

## Don't dodge "is/are" and other plain verbs

Don't replace simple state-of-being or action verbs with ornate substitutes that add no real precision.

Avoid: "Silence pooled in the room." / "The room breathed a hush."
Prefer: "The room was quiet." If the plain verb accurately describes what's happening, use it. Reach for something else only when it adds genuine specificity, not just texture.

## Avoid the "not just X, but Y" construction

This rhetorical formula is fine used once in a while, but becomes a noticeable tic when repeated. Vary sentence structure instead of leaning on this pattern for emphasis.

Avoid: "It wasn't just a house — it was a memory." Two paragraphs later: "She wasn't just tired; she was hollowed out."
Prefer: State the stronger claim directly. "The house held every memory she had." / "She was hollowed out."

## Avoid mechanical rule-of-three lists

Don't default to listing three near-synonymous adjectives or clauses for cadence. If you use a list of three, each item should do distinct work — escalate, contradict, or specify — not just restate the same idea three ways.

Avoid: "She felt fear, dread, and a creeping sense of unease." / "The city was loud, chaotic, and alive with noise."
Prefer: "She felt fear." Or, if escalation is warranted: "First fear. Then, when he didn't answer, something worse."