name-revision

Name Revision Skill

Revise AI-generated text by replacing overused names with creative, distinctive alternatives.

Core Problem

LLMs have predictable name preferences. The same names appear constantly across outputs: Seraphina, Lyra, Kira, Aldric, Theron, Shadowmere. These names signal "AI-generated" and break immersion.

Revision Workflow

1. Scan the text for all proper nouns (character names, place names, faction names)
2. Check each name against the banned list in references/banned-names.md
3. Flag names that appear on the list or match banned patterns
4. Generate replacements using techniques from references/naming-patterns.md
5. Replace consistently throughout the document
6. Verify no banned names remain

Replacement Rules

When generating replacement names:

• Never use any name from the banned list, even as a base
• Never use names ending in -ine, -ina, -ara, -yra, -iel, -ius for protagonists (overused suffixes)
• Avoid names that "sound fantasy" in the generic sense (elvish-adjacent, soft consonants, excessive vowels)
• Prefer names with harder consonants, unexpected combinations, real-world etymological roots
• Match the cultural/linguistic feel of the setting if established
• Maintain gender coding if present in original (or explicitly note if neutralizing)

Quick Reference: Red Flag Patterns

Names likely AI-generated if they:

• End in -phina, -sera, -lith, -wyn, -ael, -iel
• Start with Syl-, Lyr-, Cel-, Ser-, Ael-, Thal-
• Contain -shadow-, -moon-, -star-, -blade-, -storm-
• Sound like "Tolkien lite" or "YA fantasy protagonist"
• Could be a D&D 5e NPC from any official module

Output Format

When revising, present changes as:

ORIGINAL → REPLACEMENT (rationale)
Seraphina → Maret (Breton origin, grounded)
Shadowmere → Kaltvass (Germanic compound, "cold-water")

Then provide the revised text with all replacements applied.

Reference Files

• references/banned-names.md — Comprehensive list of names to never use
• references/naming-patterns.md — Linguistic techniques for original names