ToolsPromptsSkillsWorkflowsBlogAbout

Avoid AI Writing — Audit & rewrite content to remove AI writing patterns

A practical skill for any AI agent to audit and rewrite content and remove AI writing patterns ('AI-isms'). Three modes: Rewrite (default, flags and rewrites with built-in second pass), Detect (flags without rewriting, shows real problems vs judgment calls), and Edit (edits prose files in place with minimal targeted changes). 112-entry word replacement table across 3 tiers + 10 Tier 3 phrases, 69 pattern categories. Optional voice profiles (casual/professional/technical/warm/blunt). Works with Claude Code, OpenClaw, Hermes, Cursor, Codex, and any agentskills.io-compatible agent.

Avoid AI Writing — Audit & rewrite content to remove AI writing patterns preview

Prompt template

You are a writing quality assistant with access to the Avoid AI Writing skill. Audit and rewrite content to remove AI writing patterns ('AI-isms').

## Three Modes

### 1. Rewrite (default)
Flags AI patterns and rewrites the text to fix them. A built-in second pass catches patterns that survived the first edit.
Returns four sections:
1. Issues found — every AI-ism identified, with text quoted
2. Rewritten version — clean version with all AI-isms removed
3. What changed — summary of major edits
4. Second-pass audit — re-reads the rewrite and catches surviving tells

### 2. Detect
Flags AI patterns without rewriting. Shows which flags are real problems vs judgment calls.
Use when: patterns might be intentional, auditing content you don't want altered, quick scan.
Returns two sections:
1. Issues found — every AI-ism identified, grouped by severity (P0/P1/P2)
2. Assessment — which flags are clear problems vs patterns that may be intentional or effective

### 3. Edit
Edits a prose file in place (via Edit tool) with minimal, targeted changes, preserving passages that are already human.
Refuses source code, configuration, and generated data (prose rewrites can corrupt structured content).
Returns an edits-made + verification report, not the full file.

## 112-Entry Word Replacement Table (3 tiers)
- Tier 1A (frequency markers): delve, tapestry, etc. — always flag, evidence about how passage was produced
- Tier 1B (clarity edits): in order to, utilize, etc. — always flag, weighted lower (wordiness fix)
- Tier 2: flag when they cluster (e.g., multiple in one paragraph)
- Tier 3: flag only at high density; 10 multi-word phrases flag on repetition or when 3+ stack

## 69 Pattern Categories
Includes: structural detection (hashtag stuffing, bare-NP bullet lists, hedge-stacked predictions), AI-tool fingerprints (placeholders, citation markup, UTM params), rhythm/uniformity checks, conversational-register tells, writer-side tests.

## Common Patterns to Fix
- Sentence structure: "It's not X, it's Y" + hollow intensifiers + hedging → direct positive statements
- Structural issues: uniform paragraphs, formulaic openings, too-clean grammar → varied length, lead with point
- Transition phrases: "Moreover," "Furthermore," "In today's [X]" → "and," "also," or restructure
- Inline-header lists: "**Speed:** Speed improved by..." → write the point directly
- Title case headings: "Strategic Negotiations And Partnerships" → "Strategic negotiations and partnerships"
- Numbered list inflation: "Here are 7 reasons why..." → cut to 2-3 that matter
- False concession: "While X has limitations, it's still remarkable" → state the real tradeoff
- Rhetorical question openers: "What if there were a better way...?" → lead with the claim
- Chatbot artifacts: "I hope this helps! Let me know if..." → remove entirely

## Optional Voice Profiles
casual / professional / technical / warm / blunt — sets how prose should sound, independent of audience context.

## Task
[Your writing task here]

How to use

  1. 1Install: `git clone https://github.com/conorbronsdon/avoid-ai-writing ~/.claude/skills/avoid-ai-writing` (Claude Code)
  2. 2Or install as plugin: `/plugin marketplace add conorbronsdon/avoid-ai-writing` then `/plugin install avoid-ai-writing@conorbronsdon-skills`
  3. 3For rewrite mode (default): ask 'Remove AI-isms from this post' or 'Make this sound less like AI'
  4. 4For detect mode: ask 'detect,' 'flag only,' 'audit only,' 'just flag,' or 'scan'
  5. 5For edit mode: ask to edit a specific prose file in place — source code/config/generated data are refused
  6. 6Skill auto-triggers from phrases like 'remove AI-isms,' 'audit this draft for AI tells,' 'make this sound less like AI'
  7. 7First pass: identify all AI-isms with quoted text, apply 112-entry word replacement table
  8. 8Second pass (built-in): re-read the rewrite, catch surviving tells (recycled transitions, lingering inflation, copula swaps)
  9. 9Apply optional voice profile if specified (casual/professional/technical/warm/blunt)
  10. 10Return structured output: issues found → rewritten version → what changed → second-pass audit

Pro tips

  • 112-entry word replacement table is not vibes-based — every flagged word has a specific, plainer alternative ('Leverage' → 'use,' 'Commence' → 'start')
  • Tier 1 splits into 1A (frequency markers, evidence about production) and 1B (clarity edits, weighted lower) — only 1A counts toward AI classification
  • Tier 3 phrases (multi-word boilerplate like 'the integration of,' 'decentralized compute') flag on per-phrase repetition or when 3+ distinct phrases stack — the LLM-self-varies-boilerplate shape
  • Two-pass detection is critical — the second pass catches patterns that survive the first edit (recycled transitions, lingering inflation, copula swaps that snuck through)
  • Detect mode distinguishes real problems from judgment calls — some patterns may be intentional or effective in context
  • Edit mode preserves passages that are already human — minimal, targeted changes only, not a full rewrite
  • Voice profiles (casual/professional/technical/warm/blunt) set how prose should sound, independent of audience context
  • Works across platforms: Claude Code, Cowork (plugin), OpenClaw (ClawHub), Cursor (.mdc rule), Hermes, Codex, Windsurf, Cline, GitHub Copilot, Claude.ai Projects, ChatGPT Custom GPTs

Use cases

Removing AI writing patterns from blog posts, articles, and marketing copyAuditing content for AI tells before publicationMaking AI-generated text sound more human and naturalEditing prose files in place with minimal targeted changesQuick scanning of content for AI patterns without rewritingImproving writing quality by removing formulaic structures and inflated languageApplying consistent voice profiles (casual/professional/technical/warm/blunt) to contentSecond-pass auditing to catch AI patterns that survived initial editingDistinguishing real AI problems from intentional stylistic choicesRemoving chatbot artifacts ('I hope this helps!', 'Let me know if...')

Examples

Rewrite blog post

Input: Remove AI-isms from this blog post about productivity

Output: Invokes rewrite mode. First pass: identifies AI-isms (delve, leverage, moreover, furthermore, title case headings, numbered list inflation, rhetorical question opener, 'It's not X, it's Y' structure). Applies 112-entry replacement table. Second pass: catches surviving tells. Returns: issues found (with quotes) → rewritten version → what changed → second-pass audit.

Detect only

Input: Audit this draft for AI tells — just flag, don't rewrite

Output: Invokes detect mode. Flags all AI-isms grouped by severity (P0/P1/P2). Assesses which flags are clear problems vs patterns that may be intentional or effective in context. Returns: issues found (grouped by severity) → assessment (real problems vs judgment calls). No rewriting performed.

Edit file in place

Input: Edit this markdown file to remove AI writing patterns

Output: Invokes edit mode. Checks file type — refuses source code, configuration, generated data (prose rewrites can corrupt structured content). For valid prose files: makes minimal, targeted changes via Edit tool, preserving passages that are already human. Returns: edits-made + verification report (not the full file).

Apply voice profile

Input: Make this sound less like AI and more casual

Output: Invokes rewrite mode with casual voice profile. Removes AI-isms while applying casual tone — more conversational, less formal, contractions allowed, simpler sentence structure. Returns structured output with the casual voice applied throughout.

Key parameters

112-entry word replacement table across 3 tiers + 10 Tier 3 phrases69 pattern categories (structural, AI-tool fingerprints, rhythm, conversational, writer-side)Three modes: Rewrite (default), Detect (flag only), Edit (in-place minimal changes)Optional voice profiles: casual / professional / technical / warm / bluntBuilt-in second-pass audit catches surviving AI tellsInstallation: git clone to ~/.claude/skills/ or plugin marketplaceWorks with: Claude Code, Cowork, OpenClaw, Cursor, Hermes, Codex, Windsurf, Cline, GitHub Copilot, Claude.ai, ChatGPTLicense: Open source (check repository for specific license)Author: Conor Bronsdon

Not this

  • Not an AI detector tool — doesn't score or classify text as AI-written, just identifies and removes patterns
  • Not a grammar checker — focuses on AI-specific writing patterns, not general grammar/spelling
  • Not for source code, configuration, or generated data — edit mode refuses these (prose rewrites can corrupt structured content)
  • Not a guarantee of undetectability — removes common patterns, but sophisticated detectors may still identify AI-assisted writing
  • Not automatically installed — requires manual installation to skills directory or plugin marketplace

Works best with

ClaudeCursorcodexopenclaw

Ratings & Reviews

Loading reviews…

Share your experience

Select a rating

Related Skills

No AI Slop — Remove 20+ AI Writing Patterns While Preserving Your Voice
Writing

No AI Slop — Remove 20+ AI Writing Patterns While Preserving Your Voice

A writing editing skill that detects and removes 20+ patterns of AI-generated slop (binary contrasts, throat-clearing openers, faux-insight setups, colon reveals, dramatic fragments, importance puffery, weasel attribution, synonym cycling, fake-profound endings, and more) without flattening your personal vocabulary, cadence, humor, and imperfections.

WritingAI Slop DetectionContent Editing
4.7
YouTube to Ebook — Convert any YouTube channel into a structured EPUB ebook
Writing

YouTube to Ebook — Convert any YouTube channel into a structured EPUB ebook

An Agent Skill that transforms YouTube channel content into well-structured EPUB ebooks. Give it a channel URL or playlist and it downloads videos, transcribes them, organizes the content into logical chapters, and produces a professionally formatted EPUB ebook. Perfect for content creators who want to repurpose video content into books, educators who want to create course materials from lectures, and learners who prefer reading over watching. Supports custom chapter organization, table of contents, cover images, and metadata.

YouTubeEbookEPUB
4.6
Ecommerce Image Skills — 12 Agent Skills for Product Photography
Design

Ecommerce Image Skills — 12 Agent Skills for Product Photography

Twelve installable Agent Skills that turn a casual product photo into platform-ready ecommerce images — Amazon main images, lifestyle scenes, infographics, A+ modules, size charts, on-model shots, sale banners, Xiaohongshu covers and Taobao detail pages. Powered by GPT Image 2.5 (Flare / Sunburst).

EcommerceProduct PhotographyAmazon
4.7