Learn My Writing Style

---
name: learn-writing-style
description: Analyzes a folder of writing samples to extract the author's distinctive
  style — tone, voice, sentence structure, vocabulary patterns, formatting habits,
  and rhetorical tendencies — then writes a comprehensive Markdown style guide to a
  specified file. Use this when the user wants Claude to write in their voice, maintain
  consistent tone across documents, or capture their writing style for future reference.
  Triggers on "learn my writing style", "analyze my writing", "write like me",
  "capture my voice", "study these samples", "what's my writing style", or any request
  to understand or replicate a personal writing style.
argument-hint: <samples-folder> [output-path]
allowed-tools: Read, Glob, Write
---

# Learn Writing Style

Analyzes writing samples from a folder and produces a Markdown style guide.

## Invocation

```
/learn-writing-style <samples-folder> [output-path]
```

- `samples-folder`: Path to a directory containing writing samples (`.md`, `.txt`, `.mdx`)
- `output-path`: Where to save the style guide (default: `~/.claude/writing-style.md`)

## Workflow

1. **Parse arguments** from `$ARGUMENTS`: extract `samples-folder` (required) and
   `output-path` (optional, default `~/.claude/writing-style.md`)
2. **Glob the folder** for `*.md`, `*.txt`, `*.mdx` files — up to 30 files
3. **Prioritize strategically**: if more than 30 files exist, prefer longer,
   more varied files; note which were skipped
4. **Read each file** — skip any single file over 10,000 words; log skipped files
5. **Analyze** across all dimensions in
   [references/analysis-framework.md](references/analysis-framework.md)
6. **Synthesize**: identify patterns that are consistent vs. variable across samples,
   and determine the 5–7 most distinctive traits
7. **Draft the style guide** using the template in
   [references/style-guide-template.md](references/style-guide-template.md)
8. **Write the style guide** to `output-path`
9. **Report** the output path, number of files analyzed, and 3–5 key style traits found

## Sampling Limits

- Max 30 files per run
- Skip any single file exceeding 10,000 words
- If the total corpus exceeds ~150,000 words, read only the first ~5,000 words of
  longer files and note the truncation
- Always list skipped or truncated files in the final report

## Supporting Files

- Analysis dimensions and how to characterize them:
  [references/analysis-framework.md](references/analysis-framework.md)
- Output structure and section template:
  [references/style-guide-template.md](references/style-guide-template.md)