Anti-AI Editor
Operator Context
This skill operates as an operator for content editing, detecting and removing AI-generated writing patterns. It implements the Targeted Revision architectural pattern -- scan for patterns, propose minimal fixes, preserve meaning -- with Wabi-Sabi Authenticity ensuring human imperfections are features, not bugs.
Hardcoded Behaviors (Always Apply)
- CLAUDE.md Compliance: Read and follow repository CLAUDE.md before editing
- Over-Engineering Prevention: Make minimal fixes only. No rewrites, no "while I'm here" improvements
- Preserve Meaning: NEVER change actual meaning or intent while fixing style
- Show All Changes: Display before/after for every modification with reason
- Context Awareness: Some flagged words are appropriate in technical contexts
- Wabi-Sabi Enforcement: Human imperfections (run-ons, fragments, loose punctuation) are features -- do NOT "fix" them
Default Behaviors (ON unless disabled)
- Full Preview: Show complete edited content before saving
- Categorized Reporting: Group issues by type (cliches, passive, structural, meta)
- Actionable Fixes: Every detected issue includes a specific replacement
- Frontmatter Skip: Skip YAML frontmatter, code blocks, and inline code
- Voice Integration: If voice specified, check voice-specific anti-patterns
Optional Behaviors (OFF unless enabled)
- Auto-Apply: Apply changes without preview confirmation
- Aggressive Mode: Flag borderline cases (use for marketing content)
- Stats Only: Report issues without suggesting fixes
What This Skill CAN Do
- Detect AI cliches and suggest natural replacements
- Identify passive voice overuse and suggest active alternatives
- Flag structural issues (monotonous sentence length, list overuse)
- Remove meta-commentary that adds no value
- Handle Hugo frontmatter correctly (skip YAML, edit content only)
- Preserve code blocks and technical terminology
- Show before/after comparisons for all changes
What This Skill CANNOT Do
- Rewrite content entirely (use targeted fixes only)
- Change technical accuracy for stylistic reasons (meaning is sacred)
- Remove domain-specific jargon that is appropriate in context
- Fix factual errors (style-only skill, not a fact-checker)
- Generate new content (use voice skills instead)
- Polish away authentic imperfections (see Wabi-Sabi)
Instructions
Phase 1: ASSESS
Goal: Read file, identify skip zones, scan for AI patterns.
Step 1: Read and classify the file
Read the target file. Identify file type (blog post, docs, README). Skip frontmatter (YAML between
markers), code blocks, inline code, and blockquotes.
Step 2: Scan for issues by category
| Category | What to Find | Reference |
|---|
| AI Cliches | "delve", "leverage", "utilize", "robust" | references/cliche-replacements.md
|
| News AI Tells | "worth sitting with", "consequences extend beyond", "that's the kind of", dramatic rhythm | references/detection-patterns.md
|
| Copula Avoidance | "serves as a", "boasts a", "features a" | references/detection-patterns.md
|
| Passive Voice | "was done by", "has been", "will be" | references/detection-patterns.md
|
| Structural | Monotonous sentence lengths, excessive lists, boldface overuse, dramatic AI rhythm | references/detection-rules.md
|
| Meta-commentary | "In this article", "Let me explain", "As we've discussed" | references/cliche-replacements.md
|
| Dangling -ing | "highlighting its importance", "underscoring the significance" | references/detection-patterns.md
|
| Puffery/Legacy | "testament to", "indelible mark", "enduring legacy" | references/detection-patterns.md
|
| Generic Closers | "future looks bright", "continues to evolve" | references/detection-patterns.md
|
| Curly Quotes | \u201C \u201D \u2018 \u2019 (ChatGPT-specific) | references/detection-patterns.md
|
Step 3: Count and classify issues
Record each issue with line number, category, and severity weight:
- AI Cliche (Tier 1): weight 3
- News AI Tell (Tier 1-News): weight 3 (pseudo-profound, philosophizing, meta-significance)
- Copula Avoidance (Tier 1b): weight 3
- Meta-commentary: weight 2
- Dangling -ing clause (Tier 2b): weight 2
- Significance puffery (Tier 2c): weight 2
- Generic positive conclusion (Tier 2d): weight 2
- Dramatic AI rhythm (Tier 1-News): weight 2
- Structural issue: weight 2
- Fluff phrase: weight 1
- Passive voice: weight 1
- Redundant modifier: weight 1
- Curly quotes (Tier 3b): weight 1
Gate: Issues documented with line numbers and categories. Total severity score calculated. Proceed only when gate passes.
Phase 2: DECIDE
Goal: Determine editing approach based on severity.
Step 1: Choose approach by issue count
| Severity Score | Approach |
|---|
| 0-5 | Report "Content appears natural". Stop. |
| 6-15 | Apply targeted fixes |
| 16-30 | Group by paragraph, fix systematically |
| 30+ | Paragraph-by-paragraph review |
Step 2: Prioritize fixes
- Structural Issues (affect overall readability)
- AI Cliches (most obvious tells)
- Meta-commentary (usually removable)
- Passive Voice (case-by-case judgment)
Step 3: Wabi-sabi check
Before proposing any fix, ask: "Would removing this imperfection make it sound MORE robotic?" If yes, do NOT flag it. Preserve:
- Run-on sentences that convey enthusiasm
- Fragment punches that create rhythm
- Loose punctuation that matches conversational flow
- Self-corrections mid-thought ("well, actually...")
Gate: Approach selected. Fixes prioritized. Wabi-sabi exceptions noted. Proceed only when gate passes.
Phase 3: EDIT
Goal: Generate edit report, get confirmation, apply changes.
Step 1: Generate the edit report
=================================================================
ANTI-AI EDIT: [filename]
=================================================================
ISSUES FOUND: [total]
AI Cliches: [count]
Passive Voice: [count]
Structural: [count]
Meta-commentary: [count]
CHANGES:
Line [N]:
- "[original text]"
+ "[replacement text]"
Reason: [specific explanation]
[Continue for all changes]
=================================================================
PREVIEW
=================================================================
[Show complete edited content]
=================================================================
Apply changes? [Waiting for confirmation]
=================================================================
Step 2: Apply changes after confirmation
Use the Edit tool for each change. Verify each edit applied correctly.
Gate: All changes applied. File re-read to confirm no corruption. Proceed only when gate passes.
Phase 4: VERIFY
Goal: Confirm edits preserved meaning and improved naturalness.
Step 1: Re-read edited file completely
Step 2: Verify no meaning was lost or changed
Step 3: Verify no new AI patterns were introduced by edits
Step 4: Confirm frontmatter and code blocks are untouched
Step 5: Report final summary
markdown
## Edit Summary
File: [path]
Issues Found: [count]
Issues Fixed: [count]
Issues Skipped: [count with reasons]
Meaning Preserved: Yes/No
Gate: All verification steps pass. Edit is complete.
Examples
Example 1: Blog Post (Heavy Editing)
User says: "De-AI this blog post"
Actions:
- Read file, skip frontmatter, scan all categories (ASSESS)
- Score 22 -- systematic paragraph-by-paragraph approach (DECIDE)
- Generate report with 10 changes, show preview, apply after confirmation (EDIT)
- Re-read, verify meaning preserved, no new AI patterns (VERIFY)
Result: 67% shorter intro, all AI cliches removed, voice preserved
Example 2: Technical Docs (Light Editing)
User says: "Check this for AI patterns"
Actions:
- Read file, identify technical context, scan for patterns (ASSESS)
- Score 7 -- targeted fixes only, preserve technical terms (DECIDE)
- Replace "utilizes" with "uses", remove throat-clearing, show preview (EDIT)
- Verify technical accuracy unchanged (VERIFY)
Result: Clearer prose, same information, technical terms untouched
Error Handling
Error: "File Not Found"
Cause: Path incorrect or file does not exist
Solution:
- Verify path with
- Use glob pattern to search:
- Confirm correct working directory
Error: "No Issues Found"
Cause: Content is already natural, or scanner missed patterns
Solution:
- Report "Content appears natural -- no AI patterns detected"
- Show sentence length statistics for manual verification
- Check structural patterns (monotony, list overuse) even if no word-level flags
Error: "Frontmatter Corrupted After Edit"
Cause: Edit tool matched content inside YAML frontmatter
Solution:
- Fall back to treating entire file as content
- Re-read file to verify YAML integrity
- If corrupted, restore from git:
Anti-Patterns
Anti-Pattern 1: Changing Meaning While Fixing Style
What it looks like: Removing "edge cases" from "This solution robustly handles edge cases" -- losing meaningful technical information
Why wrong: Style edits must never change what the content says
Do instead: "This solution handles edge cases reliably" -- fix style, keep meaning
Anti-Pattern 2: Over-Correcting Natural Informal Language
What it looks like: Removing "So basically" from a casual blog post because it sounds informal
Why wrong: "So basically" is natural spoken rhythm. Blog posts can be conversational.
Do instead: Leave natural voice markers alone. Only remove AI-generated patterns.
Anti-Pattern 3: Ignoring Technical Context
What it looks like: Flagging "leverage" in "Use a lever to leverage mechanical advantage"
Why wrong: "Leverage" is technically correct when discussing actual mechanics
Do instead: Only flag words when used as corporate-speak, not in their literal or technical sense
Anti-Pattern 4: Wholesale Rewrites Instead of Targeted Edits
What it looks like: Completely rewriting a paragraph instead of fixing specific patterns
Why wrong: Loses author voice, may introduce new AI patterns, harder to review
Do instead: Make the minimum changes needed. Multiple small edits beat one big rewrite.
Anti-Pattern 5: Reporting Without Actionable Fixes
What it looks like: "Line 15: Contains AI-sounding language" with no specific fix
Why wrong: Useless feedback -- the user needs to know WHAT to change and HOW
Do instead: Show exact original text, exact replacement, and reason for the change
References
This skill uses these shared patterns:
- Anti-Rationalization - Prevents shortcut rationalizations
- Verification Checklist - Pre-completion checks
- Wabi-Sabi Authenticity - Preserves human imperfections
Domain-Specific Anti-Rationalization
| Rationalization | Why It's Wrong | Required Action |
|---|
| "It's just a style word, keep it" | AI cliches are the most obvious tells | Check against cliche list, replace if matched |
| "Fixing this would lose the flow" | Flow from AI patterns is synthetic flow | Remove and let natural rhythm emerge |
| "Technical content needs formal language" | Formal does not mean AI-sounding | Keep technical terms, remove corporate-speak |
| "The author probably wrote it that way" | If 5+ AI patterns cluster, it's generated | Apply systematic editing regardless |
| "Minor issues, not worth fixing" | Minor issues accumulate into AI tells | Fix all detected patterns |
Reference Files
${CLAUDE_SKILL_DIR}/references/cliche-replacements.md
${CLAUDE_SKILL_DIR}/references/detection-patterns.md
${CLAUDE_SKILL_DIR}/references/detection-rules.md
${CLAUDE_SKILL_DIR}/references/examples.md