Will edits to a skill break when the original is updated?

No — installed skills are local copies. Editing your local copy doesn't affect the original, and updates to the original don't overwrite your edits. To pick up updates, you'd manually merge changes, just like any text file.

What brand information is most valuable to add?

The highest-impact additions are: your brand voice (tone, vocabulary, what to avoid), your product's key differentiators and target audience, competitor context, and any hard constraints like required CTAs or compliance language.

Should I edit the skill file directly, or add a separate brand file?

Both work. Editing inline is cleaner for a single skill. For brand context you want applied across many skills, create a separate brand.md file in your .claude/ folder — the AI reads all files in that directory automatically.

How to edit a skill for your brand

Every skill in the directory is a starting point. The real value comes when you add your context — your brand voice, your product, your audience — so the AI produces output that sounds like you, not a generic marketing bot.

Why this matters

A skill gives your AI the framework. Your edits give it the context.

Without brand context, even a great skill produces output you have to rewrite. With it, you get first drafts you can actually use.

The three things to add

1. Brand voice

Add a voice section to any skill that produces written content:

## Brand voice

**Tone:** Direct, confident, never salesy. We write like a smart colleague,
not a marketer.

**Vocabulary:**
- Use: "teams", "people", "work"
- Avoid: "stakeholders", "leverage", "synergy", "game-changing"
- Product name: always "Acme" (never "the Acme platform" or "Acme's solution")

**Sentence style:** Short sentences. Active voice. No filler phrases.
("We help teams ship faster" not "Our platform enables teams to achieve
greater velocity in their shipping processes.")

**Examples of our voice:**
- ✓ "Ship faster. Break less. Sleep better."
- ✗ "Accelerate your development velocity while maintaining quality standards."

2. Product and audience context

Add a context section with the specifics your AI needs:

## Our product and audience

**Product:** [Your product name] — [one sentence description]
**Category:** [What category you compete in]
**Differentiator:** [What makes you different from alternatives]

**Primary audience:** [Job title] at [company type], usually [company size].
They care about [top 2-3 concerns]. They're skeptical of [common objections].

**Current stage:** [Early-stage / Growth / Enterprise]
**Primary acquisition channel:** [How most customers find you]

3. Examples of good output

Nothing calibrates an AI faster than showing it what "good" looks like for your specific brand:

## Example outputs

**Good subject line for us:**
"How Figma reduced onboarding drop-off by 34%"
(specific company, specific metric, no hype)

**Bad subject line for us:**
"Boost your user activation with these proven strategies"
(generic, sounds like everyone else)

**Good CTA:**
"See how it works →"

**Bad CTA:**
"Start your free trial today and experience the difference!"

How to edit safely

Skills are markdown files — edit them directly. A few rules:

Add, don't replace. Keep the original framework sections. Add your brand context as new sections at the top or bottom.

Version it. Keep the original file and your edited version. If the original skill gets updated by its author, you can diff and merge.

Name it clearly. If you fork email-sequence.md, save your version as email-sequence-acme.md so you know which is which.

Editing for a specific campaign

For one-off campaigns, you don't need to edit the skill file. Just add context in your prompt:

"Using the email-sequence skill, write a 3-email re-engagement sequence. Our product is [X], our audience is [Y], and we last emailed them [Z weeks ago] about [topic]."

Edit the skill file when you want the context to persist across all future sessions. Use the prompt for one-time specifics.

What not to edit

Leave the core frameworks alone. The reason the skill works is because of the expert thinking baked into the frameworks and decision trees. If you remove those and just add your brand context, you're writing a prompt — not improving a skill.

Add your context on top of the existing structure. Don't replace it.