cyberneticlibrary

Build reusable brand voice profiles

brand-voiceskillsetup L2182,740
affaan-m/everything-claude-code

Causal-lift measurements

brand-voice-management0pp vs no-skill baselinewith-skill 100% · baseline 100%

Measured by running the task with and without this artifact, K=5, graded by deterministic checks — no LLM judging.

What it does

Build reusable brand voice profile from real source material

Best for

Content teams, founders, or agencies that need ONE voice standard to apply across channels (X, LinkedIn, email, blog, outreach) without re-deriving style for each task.

Inputs
  • · 5-20 representative content samples (X posts, articles, essays, emails, product docs)
  • · Content age range (recent + older to detect evolution)
  • · Voice context (public launch vs. private working docs separation if exists)
Outputs
  • · VOICE PROFILE structure (reusable across downstream skills)
  • · Tone/rhythm/compression/capitalization/parenthetical/question/claim/transition norms
  • · What the author never does (banned patterns)
Requires
  • · WebFetch or firecrawl (optional, for live content retrieval)
  • · X API (optional, for live recent posts)
Preconditions
  • · 5+ representative samples collected (prefer recent over old unless user says older is canonical)
Failure modes
  • · Over-interprets single anomalous post
  • · Analyzes outdated voice (brand evolves, profile stales)
  • · Merges 'launch voice' and 'working docs voice' without separation
  • · Generic literary criticism instead of operational reuse
Trust signals
  • · Source priority ranking (recent original X posts > articles > outbound emails > product docs)
  • · Collection rules (5-20 samples, separate public vs. private if both exist, mix of topics and recency)
  • · Voice dimensions to extract (rhythm, compression, capitalization, parentheticals, questions, sharpness, numbers/mechanisms, transitions, what is never done)
  • · Schema reference to voice-profile-schema.md for structured output