| name | humanizer |
| description | Editor pipeline for humanizing content. ALWAYS USE before publishing any essay, blog post, or video script. Applies voice guides, AP Style, AI trope detection, and Writer.com review. Pipeline: (1) AI detection, (2) trope scan, (3) humanize + voice, (4) style check, (5) output.
|
| disable-model-invocation | true |
| allowed-tools | Bash(humanizer:*) |
Humanizer Skill
Purpose
Blog posts generated by AI can sound robotic, formal, or overly structured. This skill humanizes them by:
- Adding conversational transitions and voice
- Reducing repetition and redundant phrases
- Introducing personal perspective where appropriate
- Varying sentence structure and length
- Adding natural discourse markers
- Ensuring references are properly formatted and cited
When to Use
Use this skill when:
- A blog post feels too formal or robotic
- Posts need references added
- Content needs a more personal, conversational tone
- Posts are being prepared for publishing
Instructions
1. Read the Post
Read the markdown blog post completely.
2. Assess the Current Tone
Identify:
- Overly formal language
- Repetitive sentence structures
- Missing conversational transitions
- Lack of personal perspective
- Missing or incomplete references
3. Humanize the Content
Apply these transformations:
Voice & Perspective
- Add personal voice: "I think...", "What struck me...", "The interesting part is..."
- Use contractions naturally: "it's", "they're", "you'll"
- Include conversational transitions: "Here's the thing...", "But here's where it gets interesting...", "What this means is..."
- Add rhetorical questions: "Why does this matter?", "What's the real story here?"
Structure
- Vary sentence length deliberately
- Break up dense paragraphs
- Use bullet lists for key points
- Add subheadings for scanability
Tone Adjustments
- Replace "It is important to note that..." with "The key point: ..."
- Replace "One must consider..." with "You might wonder..."
- Replace "In conclusion..." with "The bottom line: ..."
- Replace passive voice with active: "The report found..." → "Anthropic found..."
Punctuation
- Remove all em-dashes (—) and replace with appropriate punctuation
- Use commas for clauses that need separation
- Use periods to split into separate sentences when it improves readability
- Never leave the gap as just a space—always use proper punctuation
4. Add References (If Missing)
Every post should have a References section at the bottom:
## References
- [Article Title](https://url) - Publication Name
- [Another Article](https://url) - Publication Name
If references are missing:
- Search for the news articles mentioned in the post
- Add proper markdown links with titles
- Include publication name when relevant
5. Preserve Key Information
While humanizing:
- Keep all factual claims accurate
- Don't change dates, numbers, or names
- Maintain the post's core message
- Keep the existing structure unless reorganizing improves readability
6. Output Format
Return the complete humanized markdown post with:
- YAML frontmatter preserved
- Humanized body content
- Properly formatted References section
Example
Before (Robotic)
The acquisition of Vercept by Anthropic represents a significant milestone in AI development. It is important to note that Claude's performance on OSWorld benchmarks has improved substantially. One must consider that computer use agents represent a paradigm shift.
After (Humanized)
When Anthropic acquired Vercept, the headline was just acquisition news. But the real story? Claude's OSWorld benchmark performance jumped from under 15% to 72.5% in a year. That's not incremental improvement—that's a step change. And it matters because "computer use" means exactly what you think: the model sees a screen, clicks buttons, types into forms, and navigates between applications. Not through APIs—through the actual interface humans use.
References
The editor pipeline uses these reference files. Load them all when humanizing:
Editor Pipeline (MANDATORY)
Always run this pipeline before publishing. No exceptions.
Step 1: AI Detection (Writer.com)
curl -X POST https://api.writer.com/v1/tools/ai-detect \
-H "Authorization: Bearer $WRITER_API_KEY" \
-H "Content-Type: application/json" \
-d '{"content": "YOUR TEXT HERE"}'
- If score > 0.7 → Content needs major humanization
- If score 0.5-0.7 → Apply trope fixes
- If score < 0.5 → Light polish only
Step 2: Trope Scan
Search for banned phrases:
- "Let's dive into" / "Let's explore" / "Let's unpack"
- "In today's landscape" / "In the ever-evolving"
- "It's important to note" / "It's worth noting"
- Em-dash count (>5 per 500 words = flag)
- Contraction ratio (<30% = too formal)
Step 3: Humanize + Voice
Apply transformations:
- Read all four reference files
- Add personal voice: "I think...", "What struck me..."
- Vary sentence length (short + long + short)
- Remove AI tropes from checklist
- Apply Scout voice guide tone rules
- Apply blog voice guide punctuation rules
- Apply AP Style guide formatting
Step 4: Style Check
Verify:
- No em dashes (use commas, periods, or restructure)
- No Oxford comma
- Numbers: 1-9 spelled out, 10+ numerals
- Times: 3 p.m. format
- Percentages: 5% symbol
Step 5: Re-Run AI Detection
Iterate until score < 0.5.
Workflow Summary
┌─────────────────────────────────────────────────────────────┐
│ EDITOR PIPELINE │
├─────────────────────────────────────────────────────────────┤
│ 1. Writer.com AI Detection → Score > 0.7? │
│ 2. Trope Scan → Flag banned phrases │
│ 3. Humanize → Apply all 4 reference guides │
│ 4. Style Check → AP Style + punctuation │
│ 5. Re-detect → Iterate until < 0.5 │
│ 6. Output → Ready to publish │
└─────────────────────────────────────────────────────────────┘
Tags
blog, writing, humanize, content, references, editor-pipeline
