skill-voice-builder · Personal Voice Profile

A skill that builds a creator's personal voice profile: a structured 6-question interview plus analysis of 3–5 writing samples produces about-me.md (personal positioning, ≤300 words) and voice.md (the voice: tone, diction, rhythm, signature phrases plus things this voice never does, ≤500 words) — so all future content sounds like you.

Analysis is strictly evidence-based: patterns are extracted from the samples, never invented, and stylistic contradictions between samples are kept as-is. No samples? Load the built-in sample set to try it out.

Example invocation: "Build my voice profile."

Full brief

Positioning

skill-voice-builder builds a creator's personal voice profile: answer 6 positioning questions, then have 3–5 real writing samples distilled into two lightweight files. The goal is that everything you create afterwards "sounds like you" instead of generic AI prose.

Core capabilities

• Two-batch interview (6 questions): identity, audience, topic pillars, industry views, persona promise, content no-go zones — the interview starts the moment the skill loads.
• Sample analysis: pattern-hunting across all samples — voice signals (sentence length, rhythm, opener style, pronouns, signature phrases), structure signals (length, list vs prose, open/close patterns), theme signals (recurring topics, stance), and absence signals (words and opener types that never appear).
• Two-file output: about-me.md (≤300 words, positioning) and voice.md (≤500 words, positive and absence signals in one file).
• Evidence discipline: every section grounded in the samples, never invented; contradictions kept, not smoothed; voice.md never repeats audience/theme info already in about-me.md.
• Profile linkage: with a profile, known answers pre-fill the interview and are skipped or verified; key points can be written back to profiles/<name>/style.md.

Workflow

1. Auto-start the interview: the first message is a question, no preamble
2. Two batches: batch 1 (identity, audience, pillars, views) → batch 2 (promise, no-go zones)
3. Generate about-me.md (≤300 words)
4. Collect 3–5 writing samples (minimum 3 to proceed; type use samples to load the built-in set)
5. Analyze across four signal dimensions
6. Generate voice.md (≤500 words, positive and absence signals together)
7. Confirm completion; future work references the profile automatically

Inputs & outputs

Output: about-me.md and voice.md in the project root (standalone lightweight artifacts, not part of the six-dimension profile schema; key points can be written back to style.md if desired).

Boundaries with adjacent skills

• skill-audience-profiler: builds the target audience profile; this skill builds the creator's own voice.
• skill-brand-onboarding: builds a brand account's full six-dimension profile; this skill is a lightweight standalone artifact solving only "write like me".

Fit

• Personal IPs wanting one consistent voice across platforms
• Teams documenting the creator's language DNA before running the account for them
• Profiles wanting a voice grounded in real samples rather than description

Before you start

• Prepare 3–5 representative writing samples (posts, scripts, drafts); the more typical, the more accurate the profile.
• No samples? Type use samples to try the built-in set; replaceable anytime.
• No API key or fees involved.

skill-voice-builder is part of the Aiglade Skill library. Invoke it from the Aiglade chat box in plain language.

← All 155 skills