author-profile

name	author-profile
description	Create author profiles via questionnaire or transcript analysis for consistent article voice

Author Profile

Create and maintain consistent author voice across all articles.

Profile Location

Stored in: .article_writer/authors.json

Schema: .article_writer/schemas/authors.schema.json

Two Ways to Create an Author

Option 1: Manual Questionnaire

Ask questions in conversational groups (2-3 at a time):

Identity

1. What name/identifier for this author? (e.g., "mwguerra")
2. Display name? (e.g., "MW Guerra")
3. Professional role(s)?
4. Years/areas of experience?
5. Expertise areas?

Languages

6. Primary writing language? (e.g., pt_BR, en_US)
7. Translation target languages?

Tone (1-10)

8. Casual (1) vs Formal (10)?
9. Neutral (1) vs Opinionated (10)?

Vocabulary

10. Terms readers know (use freely)?
11. Terms to always explain?

Style

12. Signature phrases?
13. Phrases to avoid?

Positions

14. Strong technology opinions?
15. Topics to stay neutral on?

Example

16. Write 2-3 sentences in your voice as example.

Option 2: Extract from Transcripts

Use Skill(voice-extractor) for transcript analysis.

If the author has recordings (podcasts, interviews, videos, meetings):

1. Prepare transcripts - Get transcription files with speaker labels
2. Run analysis:

   bun run "${CLAUDE_PLUGIN_ROOT}"/scripts/voice-extractor.ts --speaker "Name" transcripts/*.txt
   

3. Review extracted data - Communication style, phrases, vocabulary
4. Add identity info - Name, role, expertise, languages (manual)
5. Merge - Combine extracted + manual data

Option 3: Combined Approach (Recommended)

Best results come from combining both:

1. Extract voice patterns from transcripts
2. Add identity/expertise info manually
3. Review and refine the merged profile

Author JSON Structure

{
  "id": "author-slug",
  "name": "Display Name",
  "languages": ["pt_BR", "en_US"],
  "role": "Senior Developer",
  "experience": "10+ years",
  "expertise": ["Laravel", "PHP", "Architecture"],
  "tone": {
    "formality": 4,
    "opinionated": 7
  },
  "vocabulary": {
    "use_freely": ["Controllers", "Middleware", "API"]
       
  
   
      
      
  
   
      
      
  
   
   
      
     
     
     
       
       
       
    
     
           
           
    
      
      
       
     
  
   

Voice Analysis Fields

When transcripts are analyzed, these fields are populated:

Field	Description
extracted_from	Transcript files analyzed
sample_count	Speaking turns analyzed
total_words	Total words in analysis
sentence_structure	Length, variety, question frequency
communication_style	Traits: enthusiasm, hedging, directness, etc.
characteristic_expressions	Frequently used phrases/fillers
sentence_starters	Common ways to start sentences
signature_vocabulary	Words that characterize the speaker

Using Voice Analysis When Writing

When writing articles, use voice_analysis data:

1. Sentence structure: Match avg_length and variety
2. Tone: Follow communication_style traits
3. Natural speech: Sprinkle characteristic_expressions naturally
4. Vocabulary: Prefer words from signature_vocabulary
5. Sentence starters: Use patterns from sentence_starters

Multi-Language Workflow

1. Article written in author's primary language (first in array)
2. After completion, translated to other languages
3. Each file named: {slug}.{language}.md

Example for author with ["pt_BR", "en_US"]:

content/articles/2025_01_15_rate-limiting/
├── rate-limiting.pt_BR.md    # Primary (written first)
└── rate-limiting.en_US.md    # Translation

Default Author

If article task doesn't specify author:

• First author in authors.json is used
• Their language settings apply
• Their voice/tone is followed

Updating Authors

Add More Transcript Data

# Analyze new transcripts for existing author
bun run "${CLAUDE_PLUGIN_ROOT}"/scripts/voice-extractor.ts \
  --speaker "Name" \
  --author-json \
  new_podcast.txt > new_analysis.json

# Merge into existing profile (manually or via command)

When to Update

• New transcript data available
• Writing style evolves
• Feedback indicates tone mismatch
• New expertise develops