How to use the Brand Voice Profiler agent skill (brand-voice)

What is an agent skill, and what does brand-voice make the agent do?

An agent skill is a folder containing a SKILL.md file with instructions that an AI agent loads when a task matches the skill's description. The brand-voice skill is one of these instructions files. Its job is to build an active voice profile from your real source material and then use that profile across content, outreach, and social work.

The skill works in a specific order. It prefers recent original X posts and threads first, then articles, essays, memos, launch notes, or newsletters, then real outbound emails or DMs that worked, and finally product docs, changelogs, README framing, and site copy. It asks for roughly 5 to 20 representative samples when they are available.

• Analyses rhythm and sentence length, compression versus explanation, capitalization, use of parentheses, question frequency, how sharply asks are made, how often numbers or receipts appear, and how transitions work.
• Notes what the author never does.
• Produces a reusable VOICE PROFILE block that downstream tasks can consume.
• Lists phrases and habits to avoid, such as fake curiosity hooks, forced lowercase, and fishing questions.

Install the brand-voice skill from the ZIP

Download the ZIP for the skill from this page and unzip it. The skill folder contains the file SKILL.md. Place the unzipped folder in one of these locations, depending on how you use your agent:

• Personal install: unzip into ~/.claude/skills/ so the skill is available across your projects.
• Project install: unzip into the .claude/skills/ folder inside a specific project, so the skill applies only there.
• Claude apps: upload the skill under Settings → Capabilities → Skills.

Trigger the skill with natural requests

You do not need a special command. Describe the writing task in plain language and include your samples. The skill's description tells the agent to use it when you want a consistent voice across content, outreach, and social workflows, or when you want to avoid common AI writing tropes.

Good triggers name the channel and give the agent material to learn from. Paste the samples directly into the chat or point the agent to files you have supplied.

• "Here are 10 of my recent posts. Build a voice profile from them, then draft three X posts about our release."
• "Using my voice profile, write a launch email for the new docs site."
• "Match this brand voice when you rewrite our README intro."

Example workflow: from samples to a drafted post

Start by supplying real writing. The skill's output is only as good as the samples, so choose pieces that sound like you at your best and that reflect the channel you are writing for. Ask the agent to build the profile first and check it before drafting.

Once the profile exists, ask for drafts. Review them against the avoid list. If a draft drifts into generic phrasing, tell the agent which sentence is off and why, and it can adjust while keeping the profile in view.

• Step one: supply samples and ask for a VOICE PROFILE.
• Step two: read the profile and correct anything that does not sound like you.
• Step three: request drafts for each channel and compare them with your own writing.

What you need before you start

The skill requires real sample texts from you, such as posts, essays, emails, or docs. It does not supply its own voice. If you have no samples, the skill falls back to a fixed example style tuned to the ECC author's direct, terse voice, which may not match you.

An optional companion is the x-api skill together with X API access. With both, the agent can pull your recent original posts automatically rather than relying on pasted text. Without them, you provide the samples yourself.

Safety notes: scripts and network access

The brand-voice skill does not run scripts and does not use the network on its own. It works from the text you give it. The only network-related path is the optional x-api integration, which needs your own X API access.

Use the skill only for your own voice, or for brands you are authorised to represent. Do not use it to impersonate someone else or to publish content under a real person's name without their consent, since this raises authenticity and reputational risk.

Tips for better results

Quality of samples matters more than quantity. Remove off-brand pieces, and separate public launch voice from private working voice if your material clearly splits. Recent samples are preferred over older ones unless you say older writing is more representative.

• Give five to twenty representative samples where possible.
• Keep public and private voices separate if they differ.
• Include current site copy or README framing when website wording matters.
• Review the avoid list and tell the agent about any phrases you never use.