runtsang/RebuttalStudio / skills/utility/stage/writing-anti-ai
writing-anti-ai
VerifiedRemove AI-generated writing patterns from rebuttal prose to make it sound natural, direct, and authentically human-authored. Use when a Stage 2 refined draft or Stage 4 follow-up response reads too formulaic, robotic, or "GPT-like". Supports academic English.
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 6.4 KB |
SKILL.md
View on GitHub---
name: writing-anti-ai
description: Remove AI-generated writing patterns from rebuttal prose to make it sound natural, direct, and authentically human-authored. Use when a Stage 2 refined draft or Stage 4 follow-up response reads too formulaic, robotic, or "GPT-like". Supports academic English.
tags: [Writing, Rebuttal, Anti-AI, Humanizer, Polish]
version: 1.0.0
source: Adapted from Claude Scholar (https://github.com/Galaxy-Dawn/claude-scholar), skills/writing-anti-ai/SKILL.md. Original authored by gaoruizhang; based on Wikipedia's "Signs of AI writing" guide (WikiProject AI Cleanup).
---
# Writing Anti-AI for Rebuttal Prose
> **RebuttalStudio Utility — Stage 2 / Stage 4 / Polish**
> Apply this skill after Stage 2 refinement or Stage 4 follow-up drafting when the output reads formulaic or over-polished. Use before the Stage 3 document compilation to ensure the final rebuttal sounds authored by a researcher, not generated by a model.
Remove predictable AI writing patterns from rebuttal text so it reads as natural, confident academic prose — not as machine-generated boilerplate.
---
## Why This Matters for Rebuttals
Reviewers are experienced academics who read hundreds of papers and rebuttals. A rebuttal that sounds AI-generated signals:
- Lack of genuine engagement with the concerns
- Boilerplate responses that don't address the specific paper
- Reduced credibility and persuasiveness
The goal is not to hide AI assistance — it's to ensure the **ideas and voice** of the actual authors come through clearly.
---
## Core Insight
LLMs predict statistically likely continuations. This produces text that applies to the widest variety of cases — which is exactly what makes it feel generic in a rebuttal. Rebuttal prose needs to be *specific* to this paper, this reviewer, and this concern.
**Core principle**: Every sentence in a rebuttal should only be true for *this* paper. If it could appear in any rebuttal, rewrite it.
---
## Quick Checklist (Apply in Order)
### 1. Cut Filler Phrases
Remove throat-clearing openers and emphasis crutches that add length without adding meaning.
**Common AI openers to remove:**
- "In order to address this concern…" → "To address this…"
- "It is worth noting that…" → (delete or restate directly)
- "It is important to highlight that…" → (delete)
- "We would like to clarify that…" → "We clarify that…"
- "Due to the fact that…" → "Because…"
**AI emphasis phrases to cut:**
- "crucial", "pivotal", "paramount", "essential" (unless truly warranted)
- "delve into", "unpack", "dive deeper"
- "It goes without saying that…" → (just say it)
### 2. Break Formulaic Structures
AI tends toward predictable patterns. In rebuttals, these appear as:
| AI Pattern | Rebuttal Fix |
|-----------|-------------|
| "It's not just X, it's also Y" (negative parallelism) | Drop it; state Y directly |
| "First… Second… Third…" for every single response | Only use numbered lists when the reviewer explicitly listed multiple concerns |
| Em-dash reveals: "X — which shows Y" | "X, which shows Y" or split into two sentences |
| "This represents a significant advancement…" | State what changed specifically |
### 3. Vary Rhythm
Check sentence length distribution in each response block:
- Three consecutive sentences of similar length? Break or merge one.
- Every paragraph ending with a confident one-liner? Vary endings.
- Mix short punchy sentences with longer elaborations.
### 4. Trust the Reviewer
Reviewers are experts. Do not over-explain or over-justify.
**Bad (patronizing):** "To ensure complete clarity, and so that the reviewer can fully appreciate our contribution, we explain in detail…"
**Good:** "To clarify: [the specific point]."
**Bad (hand-holding):** "It could potentially be argued that the metric might perhaps be considered somewhat limited in scope."
**Good:** "The metric has a known limitation: [state it]. We address this in Appendix B."
### 5. Remove Pull-Quote Language
If a sentence sounds like it was written to be quoted or highlighted, rewrite it.
**Bad:** "This represents a paradigm shift in how the community will approach [problem]."
**Good:** "The method reduces [specific metric] by X% over [specific baseline], enabling [specific application]."
### 6. Use First-Person Plural Correctly
Academic rebuttals use "we" — but AI-generated text often over-hedges or over-personalizes.
**Too hedged:** "The authors feel that it may be the case that…" → "We argue that…"
**Over-personalized:** "We personally believe that…" → "We believe that…"
---
## Common AI Vocabulary in Rebuttals to Avoid
These words are statistically overrepresented in AI-generated academic text:
| Overused | Replace With |
|----------|-------------|
| additionally | also / and / (restructure sentence) |
| furthermore | (restructure sentence) |
| notably | (delete or be specific) |
| it is worth noting | (just state the point) |
| comprehensive | thorough / complete (or be specific) |
| leverage | use |
| utilize | use |
| facilitate | help / enable |
| demonstrate | show |
| novel | (be specific about what is new) |
---
## Scoring: Is the Rebuttal Prose Human?
Rate each response block (0–10 points each):
| Dimension | 0 pts | 10 pts |
|-----------|-------|--------|
| **Directness** | Opens with filler, buries the answer | States the answer in the first sentence |
| **Specificity** | Could appear in any rebuttal | References this specific paper, reviewer comment, or result |
| **Rhythm** | All sentences same length | Mix of short and long sentences |
| **Tone** | Defensive or overly formal | Collegial, confident, factual |
| **Density** | Padded with acknowledgments | Information-dense; every sentence moves forward |
**40–50**: Ready to submit. **30–39**: Needs targeted edits. **< 30**: Rewrite from outline.
---
## What NOT to Change
This skill only removes AI patterns. Do **not**:
- Alter technical claims or experimental results
- Remove citations or evidence
- Change `{{placeholder}}` tokens
- Modify quoted reviewer text
- Change the structure required by Stage 2 output format (the `> **Reviewer's Comment**:` / `**Response**:` block)
---
*Adapted from Claude Scholar's `writing-anti-ai` skill. Original authored by gaoruizhang. Based on [Wikipedia: Signs of AI writing](https://en.wikipedia.org/wiki/Wikipedia:Signs_of_AI_writing), maintained by WikiProject AI Cleanup. Source: https://github.com/Galaxy-Dawn/claude-scholar/blob/main/skills/writing-anti-ai/SKILL.md*
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| slack-gif-creatoranthropics/skillsKnowledge and utilities for creating animated GIFs optimized for Slack. Provides constraints, validation tools, and animation concepts. Use when users request animated GIFs for Slack like "make me a GIF of X doing Y for Slack." | Stars: 198k | Updated | Claude CodeCodex | Verified | Install |
| ponytailDietrichGebert/ponytailLazy senior dev mode for any coding task: the smallest change that fully solves it, and a reply a busy human understands in one read. | Stars: 194k | Updated | Claude CodeCodex | Verified | Install |
| ponytail-debtDietrichGebert/ponytailHarvest every shortcut: comment into one debt ledger, so deferrals get tracked instead of forgotten. One-shot report. | Stars: 194k | Updated | Claude CodeCodex | Verified | Install |
| ponytail-debtDietrichGebert/ponytailList every `shortcut:` comment (and older `ponytail:` ones) as a debt ledger. One-shot report, changes nothing. Use for "ponytail debt", "what did ponytail defer", "list the shortcuts", /ponytail-debt. | Stars: 194k | Updated | Claude CodeCodex | Verified | Install |