Tone Mapping

Tone Mapping

Build a voice and tone profile by analyzing real writing samples. The output is brain/voice-and-tone.md, which content-producing skills read to match the user's natural voice.

Why This Matters

Without a tone map, every skill produces generic-sounding output. With one, blog posts, social content, emails, and landing pages start sounding like the person who runs the company wrote them.

Workflow

Step 1: Collect Writing Samples

Ask the user:

I need 3-5 real writing samples to map your voice. These should be things you wrote yourself (not edited by someone else) that sound most like you.

Options:

1. Paste directly — copy text from LinkedIn posts, blog posts, emails, newsletters
2. Link to a URL — I'll fetch the content (LinkedIn articles, blog posts, personal site)
3. Point to a file — if you've dropped samples in /incoming/

Which works best?

Minimum: 3 samples, ideally 500+ words total. More samples produce a better map.

If they have fewer than 3 samples: Accept what they have but flag that the profile will be approximate. Suggest they update it later as they produce more content.

What makes a good sample:

• Written by them, not ghostwritten or heavily edited
• Represents how they want to sound going forward (not how they used to write)
• Variety helps: a casual post, a technical explanation, an opinionated take

Step 2: Extract Patterns

Analyze the samples across eight dimensions. For each dimension, identify the specific pattern with evidence (quote a short phrase from the sample).

The Eight Dimensions

1. Sentence rhythm

   • Average sentence length (short/medium/long)
   • Variation pattern (do they mix short punchy sentences with longer ones, or stay consistent?)
   • How they open paragraphs (statement, question, anecdote, data?)

2. Vocabulary register

   • Formal to casual spectrum (1-5 scale)
   • Technical density (do they use jargon freely, explain it inline, or avoid it?)
   • Industry-specific terms they default to

3. Structural patterns

   • How they organize arguments (top-down conclusion first? build-up? problem-solution?)
   • Use of lists vs. prose paragraphs
   • Paragraph length tendencies

4. Pronouns and perspective

   • First person ("I") vs. collective ("we") vs. second person ("you")
   • How they address the reader (directly? as a peer? as a student?)
   • Do they use "one" or passive voice to generalize?

5. Hedging vs. conviction

   • Do they qualify claims ("might," "could," "tends to") or state directly?
   • How they handle uncertainty (acknowledge it, skip past it, frame it as a question?)
   • Strength of opinionated statements

6. Humor and personality markers

   • Dry wit, self-deprecation, sarcasm, none?
   • Use of analogies and metaphors (frequent, rare, specific domain?)
   • Personality that comes through (irreverent? earnest? measured?)

7. Punctuation and formatting habits

   • Em dash usage (heavy user? never?)
   • Parenthetical asides (frequent, rare?)
   • Bold/italic patterns
   • How they use colons, semicolons, ellipses

8. Opening and closing patterns

   • How they start pieces (hook, question, bold claim, anecdote?)
   • How they end (CTA, summary, open question, provocative statement?)
   • Transition style between sections

Step 3: Confirm with Sample Sentences

Before writing the profile, test it. Generate 3 sentences that apply the extracted patterns to a topic the user hasn't written about. Present them:

Here's how your voice profile would sound on a new topic. Do these sound like you?

1. [sentence applying their rhythm + vocabulary + conviction level]
2. [sentence applying their humor/personality + structural pattern]
3. [sentence applying their opening style + pronoun usage]

What's off? Too formal? Too casual? Missing something?

Iterate until the user confirms the voice feels right. This is the most important gate in the process.

Step 4: Write voice-and-tone.md

Write the profile to brain/voice-and-tone.md using the template structure. Include:

• The eight dimension scores with evidence
• A "Voice DNA" summary (3-4 sentences capturing the overall feel)
• A "Do / Don't" quick-reference table
• 3 sample sentences as calibration anchors

Step 5: Update references

If brain/INDEX.md exists, check whether it already lists voice-and-tone.md. If not, add it under the Brand section.

Tell the user:

Voice profile saved to brain/voice-and-tone.md. Content skills (blog, social-content, copywriting, copy-editing, email-sequence, case-studies) read it when drafting and editing.

To update your voice profile later, run /tone-mapping again with new samples.

Step 6: Offer Calibration Samples (voice-samples.md)

The profile captures the measurable dimensions of the voice. Calibration samples capture what dimensions can't: signature moves, argument structures, and real quoted examples. Offer this step:

Want me to also build brain/voice-samples.md? It's the positive companion to the profile — your signature moves with real quotes from your writing, so drafts don't just match your metrics but reuse your actual patterns. Works best with 5+ samples.

If yes, extract from the same samples (plus any new ones):

1. The voice in one paragraph — describe the voice the way you'd brief an editor
2. Signature moves (3-7) — recurring patterns that make the voice recognizable (e.g., analogy-first framing, parenthetical asides, self-deprecation, direct address). For each: what it is, 2-3 quoted examples taken verbatim from the samples, and how often to use it
3. Structural patterns — how pieces open and close, paragraph and subhead habits, with quoted examples
4. Argument structures — how they build a case (e.g., consensus-then-pivot, experience escalation), with the beats spelled out
5. What this writer rejects — patterns that never appear in their writing, stated concretely
6. Specificity calibration — how concrete the voice gets (names, numbers, tools vs. pattern-level)

Write to brain/voice-samples.md following the template structure. Confirm the signature moves with the user — show each move with its quotes and ask "is this actually you, or an artifact of one post?" Drop anything that appears in only one sample.

Hard rule: every example in voice-samples.md must be quoted verbatim from the user's real writing. If you can't quote it, don't include the move.

If brain/INDEX.md exists, list voice-samples.md there too.

Rules

• Never fabricate sample quotes. Only quote from the user's actual writing samples.
• The voice profile is descriptive, not prescriptive. Capture how they actually write, not how a style guide says they should write.
• If samples conflict (one is casual, another formal), ask which register they want as their default and note the range.
• Don't over-index on a single sample. Patterns should appear across multiple samples to be included.
• If the user already has a voice-and-tone.md, show what would change and confirm before overwriting.
• The profile should be concise enough that skills can read it quickly. Target 150-200 lines max. No padding.