Skip to main content
Impeccable is designed to adapt to your project’s specific needs. This guide covers how to customize the skills, add project context, and make Impeccable work better for your team.

Project Context with /teach-impeccable

The most important customization is establishing project-specific design context. The /teach-impeccable command is designed for exactly this purpose.

Why Project Context Matters

Without context, AI makes assumptions based on generic training data. This leads to:
  • Generic design choices (the “AI slop” aesthetic)
  • Decisions that don’t match your brand or users
  • Wasted time fixing inappropriate suggestions
With context, every command understands:
  • Who your users are and what they need
  • Your brand personality and aesthetic direction
  • Design principles specific to your product
  • What to avoid (anti-references)

Running /teach-impeccable

1

Run the command

Execute /teach-impeccable in your AI tool. This only needs to be done once per project.
Codex CLI users: Use /prompts:teach-impeccable
2

Let it explore your codebase

The command will automatically:
  • Read your README and documentation
  • Analyze package.json and dependencies
  • Review existing components and patterns
  • Examine design tokens and CSS variables
  • Find brand assets and style guides
This helps it understand what you’ve already established.
3

Answer focused questions

Based on what it found (or didn’t find), you’ll be asked targeted questions:Users & Purpose:
  • Who uses this? What’s their context?
  • What job are they trying to get done?
  • What emotions should the interface evoke?
Brand & Personality:
  • How would you describe the brand in 3 words?
  • Any reference sites that capture the right feel?
  • What should this NOT look like?
Aesthetic Preferences:
  • Visual direction? (minimal, bold, elegant, playful, etc.)
  • Light mode, dark mode, or both?
  • Any colors that must be used or avoided?
Accessibility:
  • Specific requirements (WCAG level, user needs)?
  • Considerations for reduced motion or color blindness?
4

Review generated context

The command synthesizes your answers into a Design Context section:
5

Context saved to config

This gets written to your config file:
  • Cursor: .cursorrules or .cursor/rules/impeccable.md
  • Claude Code: .claude/rules/impeccable.md
  • Gemini CLI: GEMINI.md
  • Codex CLI: .codex/rules/impeccable.md
Now every command automatically has this context!

Example: Complete Design Context


Customizing the Frontend-Design Skill

The core frontend-design skill can be customized by adding project-specific guidance.

Adding Custom Anti-Patterns

If your team has specific patterns to avoid, document them in your config:

Adding Custom Design Tokens

Document your design tokens so commands reference them:

Adding Component Patterns

Document established patterns that commands should follow:

Customizing Individual Commands

You can influence command behavior through your config file.

Example: Custom /audit Criteria

Add project-specific audit criteria:

Example: Custom /polish Standards

Define your polish standards:

Tool-Specific Customization

Each AI tool has different capabilities for customization.

Cursor

Location: .cursorrules or .cursor/rules/ Capabilities:
  • Add project context to .cursorrules
  • Create skill-specific rule files in .cursor/rules/
  • Rules are always loaded automatically
Example structure:

Claude Code

Location: .claude/rules/ or .claude/skills/ Capabilities:
  • Rich frontmatter with metadata
  • Modular rule files
  • Can reference external documentation
Example structure:

Gemini CLI

Location: GEMINI.md (root level) or .gemini/ Capabilities:
  • Import system (@./filename.md)
  • Modular documentation
  • TOML commands with custom metadata
Example structure:
GEMINI.md example:

Codex CLI

Location: .codex/ directory Capabilities:
  • Custom prompts with arguments
  • Routing to specialized agents
  • Skills system
Example structure:

Best Practices for Customization

Don’t manually write design context. Let /teach-impeccable guide you through the process and generate structured context.
Start with basic context and add more as patterns emerge. Don’t try to document everything upfront.
Focus on what makes your project unique. Don’t document things that are already standard practice.Good: “We use monospace only for code blocks, not as aesthetic choice”Bad: “Use good typography” (too vague)
Saying what NOT to do is often more valuable than saying what to do.“Anti-references: Notion (too playful), Linear (too minimal), Salesforce (too dense)”
Design context evolves. Update your config as you discover what works and what doesn’t.
Commit config files to version control. Everyone should have the same design context.

Advanced: Creating Custom Commands

You can create your own commands that extend Impeccable.

Example: Custom /audit-accessibility Command

Cursor (.cursor/commands/audit-accessibility.md):
Claude Code (.claude/commands/audit-accessibility.md):
Gemini CLI (.gemini/commands/audit-accessibility.toml):

Example: Custom /brand-check Command

Check if designs match brand guidelines:

Examples by Project Type


Troubleshooting Customization

Problem: Commands give generic suggestions despite having design context.Solutions:
  • Verify context is in correct config file location
  • Restart your AI tool to reload config
  • Make context more specific (avoid vague statements)
  • Reference context explicitly: “Check Design Context section in config”
Problem: Command runs but doesn’t write to config file.Solutions:
  • Check file permissions in project directory
  • Verify correct config file path for your tool
  • Look for existing Design Context section (might be updating)
  • Check if tool’s working directory is project root
Problem: Created custom command but tool doesn’t see it.Solutions:
  • Verify file is in correct commands directory
  • Check file name matches command name
  • Restart AI tool to reload commands
  • For Gemini CLI, check TOML syntax is valid
  • For Codex CLI, use /prompts: prefix
Problem: Project context contradicts Impeccable’s design principles.Solutions:
  • Be explicit about overrides: “For this project, ignore guideline about…”
  • Use “Project-Specific” headers to clarify precedence
  • Consider if Impeccable is the right tool (might need different approach)

Next Steps

Commands Overview

Complete reference of all 17 commands

Best Practices

When to use which commands and workflow recommendations