setup-language

Setup My Language Style

You are orchestrating an interactive session to build a personalized writing and documentation style skill. Analyze the developer's existing written content, interview them about voice preferences, then generate a skill at ~/.claude/skills/my-language/ using skill-creator.

Core Principles

• Read before profiling: Extract voice patterns from actual writing samples first
• Context-aware tone: Docs, articles, and commit messages each have different tone rules
• Capture the voice, not grammar rules: HOW the person communicates, not generic writing advice
• Preserve authenticity: Codify the developer's natural voice, not an idealized version

---

Phase 1: Discovery

1. If $ARGUMENTS is a path → use as first sample source
2. Ask for writing samples across categories:
   • "Share 1-3 sources that represent your writing voice"
   • Accepted formats:
     • Article URLs: Published blog posts, Medium articles, dev.to posts
     • Documentation paths: README files, docs/ directories, wiki pages
     • File paths: Markdown files, technical guides, tutorials
     • Git repos: Extract commit messages and PR descriptions automatically
3. If no samples provided → skip to Phase 3 (pure interview mode)
4. Verify paths via Bash (test -f or test -d). For URLs → fetch via Bash (curl)

---

Phase 2: Sample Analysis

1. For each sample source (max 3), use Glob, Grep, and Read directly in the main context — no subagents. Extract:
   • Voice traits: Personal/formal, active/passive, sentence length average
   • Opening patterns: How sections/articles/docs begin
   • Transition phrases: Recurring connectors and segue patterns
   • Code introductions: How code blocks are introduced and followed up
   • Closing patterns: Abrupt, summary, call-to-action, or friendly
   • Signature expressions: Recurring phrases, verbal tics, characteristic word choices
   • Structure patterns: Section flow, heading hierarchy, list usage, callout usage
   • Tone differences: How tone shifts between documentation vs. articles vs. comments
   • Rhetorical devices: Questions, analogies, humor usage
   • Formatting habits: Bold/italic usage, table frequency, emoji usage
2. Git availability check: run git --version to confirm git is installed, then git -C <path> rev-parse --git-dir to confirm the path is a git repo. If unavailable → skip commit/PR pattern extraction and note: "Git history unavailable — skipping commit style analysis."
3. If a git repo is provided, additionally extract:
   • Commit message style (conventional commits? imperative? past tense?)
   • PR description patterns
   • Code comment voice
4. Synthesize findings into a voice profile with direct quotes from the samples
5. Present: "Here's your writing voice profile — confirm what's accurate and flag what to adjust"