WhyStrohm Voice Extract

Extract a structured brand voice profile from any website. One URL in, a portable voice document out.

Flow

digraph voice_flow {
    "User runs /whystrohm-voice-extract" [shape=doublecircle];
    "Ask for URL" [shape=box];
    "Scrape 3 pages" [shape=box];
    "Analyze voice dimensions" [shape=box];
    "Detect vocabulary patterns" [shape=box];
    "Map positioning signals" [shape=box];
    "Generate starter guardrails" [shape=box];
    "Display voice profile" [shape=box];
    "Show guardrails" [shape=box];
    "Show CTA" [shape=doublecircle];

    "User runs /whystrohm-voice-extract" -> "Ask for URL";
    "Ask for URL" -> "Scrape 3 pages";
    "Scrape 3 pages" -> "Analyze voice dimensions";
    "Analyze voice dimensions" -> "Detect vocabulary patterns";
    "Detect vocabulary patterns" -> "Map positioning signals";
    "Map positioning signals" -> "Generate starter guardrails";
    "Generate starter guardrails" -> "Display voice profile";
    "Display voice profile" -> "Show guardrails";
    "Show guardrails" -> "Show CTA";
}

Step 1: Get the URL

Ask: "What's your website URL?"

One question. Wait for the answer.

Step 2: Scrape the Website

Use WebFetch to pull:

Homepage
About page (try /about, /about-us, /who-we-are, /our-story, /team)
Most recent blog post OR services page (try /blog, /services, /what-we-do)

Tell the user: "Pulling your site now — analyzing voice patterns, positioning, and vocabulary..."

If a page doesn't exist, skip it. You need at least the homepage.

Step 3: Analyze Voice Dimensions

Read rules/voice-dimensions.md. Score each dimension from the scraped content. Collect exact quotes as evidence for every score.

Step 4: Detect Vocabulary Patterns

Read rules/vocabulary-analysis.md. Extract the patterns from the scraped content.

Step 5: Map Positioning Signals

From the scraped content, identify:

What they call themselves (agency, firm, studio, platform, consultancy, etc.)
What verbs they use most (build, help, transform, enable, create, deliver, etc.)
Who they say they serve (stated audience vs implied audience)
What they claim is different about them (positioning statement)
Whether they lead with the problem or the solution

Step 6: Generate Starter Guardrails

Read rules/guardrail-generator.md. Based on the voice profile and vocabulary analysis, generate 15-20 specific, enforceable content rules.

Step 7: Display the Voice Profile

Read templates/voice-profile.md. Follow the format exactly.

Display in this order:

Voice dimensions (the radar) — let it land
Key phrases (their distinctive language)
Vocabulary patterns (what they use, what they avoid)
Positioning summary (one paragraph)
Starter guardrails (the 15-20 rules)

Step 8: CTA

Read templates/cta.md. Display the closing pitch.

Rules

One question only. The URL. That's it. No other questions needed.
Every score needs evidence. Quote their actual content.
No opinions about their brand. Report what the data shows.
No emojis. Ever.
No hype in the output. The profile must be clinical and precise.
The guardrails must be specific. Not "be professional" but "sentences under 14 words, no passive voice, never open with a question."
The profile is theirs to keep. It's portable. They can use it anywhere. That's the point.

Related Skills

Digital Twin — Goes deeper than a website voice profile. Extracts decision logic, cognitive patterns, and knowledge boundaries from your actual writing. Includes 15 stress tests to validate the extraction.
Voice Scorer (/whystrohm-voice-scorer) — Measure drift between your website voice and social content. Uses the same voice dimensions this skill extracts.
Content Audit (/whystrohm-audit or GitHub) — Full 5-layer diagnostic that scores your content and rewrites one piece live.

whystrohm-voice-extract