Shared Context

Shared Context Setup

Create a centralized context folder that all skills and agents reference for consistent output.

Modes

• brand — Voice profile, tone, banned words, audience personas, content pillars
• domain — Project conventions, terminology glossary, domain knowledge
• full (default) — Both brand and domain

Process

Step 1: Assess What Exists

• Check for .claude/shared-context/, .claude/brand-context/, or existing voice/persona files
• Read CLAUDE.md, README, any marketing or docs directories
• Report what's found and what's missing

Step 2: Interview

Brand questions (skip in domain mode):

1. Brand personality in 3 adjectives?
2. Primary audience: job title, pain points, language they use?
3. Words/phrases to NEVER use? (AI artifacts, competitor names, off-brand jargon)
4. Platform differences? (LinkedIn vs docs vs email vs social)
5. Share 1-2 examples of content you're proud of

Domain questions (skip in brand mode):

1. Project domain? (fintech, healthcare, e-commerce, etc.)
2. Project-specific terminology that's often misused?
3. Non-obvious conventions from reading the code?
4. Who consumes this project's output? (engineers, end users, both)

Step 3: Create Structure

.claude/shared-context/
├── README.md               # What this folder is, how to use it
├── brand/
│   ├── voice-profile.md    # Tone, banned words, platform rules, examples
│   ├── audience-personas.md # ICP with pain points and values
│   └── content-pillars.md  # Core topics (content projects only)
└── domain/
    ├── conventions.md       # Project-specific rules and gotchas
    └── glossary.md         # Domain terminology

CHECKPOINT: Show each document before writing. User confirms or edits.

Step 4: Wire Into Skills

• Scan .claude/skills/ for writing/content skills
• Offer to add Read .claude/shared-context/brand/voice-profile.md before generating content to relevant skills

Step 5: Register in CLAUDE.md

Add:

## Shared Context
Brand voice, audience personas, and domain conventions: `.claude/shared-context/`
Content-generating skills should read relevant files before producing output.