AI Context Plugin for Claude Code

Generate and manage .ai-context files for AI-assisted development.

This plugin helps you create structured context files that help AI assistants understand your project deeply - including conventions, domain terminology, architecture, and sensitive areas.

The Problem

AI assistants often struggle with project-specific knowledge:

• They don't know your naming conventions
• They miss domain-specific terminology
• They can't see your architectural decisions
• They don't know which areas need extra care

The Solution

The AI Context Plugin creates and maintains .ai-context files - structured YAML documents that capture everything an AI needs to know about your project:

• Project metadata - Stack, framework, dependencies
• Domain knowledge - Terms, entities, relationships
• Conventions - Naming patterns, file organization
• Preferences - Tooling choices, patterns to avoid
• Caution areas - Sensitive code with severity levels
• Testing - Frameworks, commands, patterns

Quick Start

# Add the marketplace
/plugin marketplace add AI-Native-Systems/ai-context-cc-plugins

# Install the plugin
/plugin install ai-context@ans

# Initialize context for your project
/ai-context:init

Commands

Command	Description
/ai-context:init	Create .ai-context for new or existing projects
/ai-context:update	Update context based on recent codebase changes
/ai-context:digest	Generate human-readable markdown summary

Skills (Auto-Triggered)

Skills are automatically invoked by Claude based on trigger phrases:

Skill	Trigger Phrases	Purpose
context-init	"context init", "create ai context", "initialize context"	Create context files
context-update	"update context", "sync context", "refresh ai context"	Keep context synced
context-digest	"context digest", "generate digest", "onboarding doc"	Generate readable summaries

How It Works

1. Context Init

For existing projects, the plugin:

1. Scans your codebase (package.json, configs, directory structure)
2. Detects stack, framework, and conventions
3. Asks you to confirm inferences
4. Fills gaps with targeted questions
5. Generates .ai-context YAML file
6. Sets up auto-loading via SessionStart hooks

For new projects, it guides you through setup questions.

2. Context Update

When your codebase evolves:

1. Detects changes via git history or file scanning
2. Identifies new dependencies, patterns, structure changes
3. Shows a diff of proposed updates
4. Makes surgical edits while preserving user-authored content

3. Context Digest

Transforms .ai-context into AI-CONTEXT-DIGEST.md:

• Human-readable markdown
• Perfect for team onboarding
• Tables, headers, and scannable format
• Skip empty sections automatically

Example .ai-context Output

version: "1.0"

project:
  name: "my-app"
  description: "E-commerce platform"
  type: "web-app"
  stack:
    - TypeScript
    - Next.js
    - Tailwind CSS
    - Prisma

domain:
  industry: "e-commerce"
  terms:
    - term: "SKU"
      meaning: "Stock Keeping Unit - unique product identifier"
    - term: "Cart"
      meaning: "Temporary collection of items before checkout"

structure:
  conventions:
    components: "src/components/{Name}/{Name}.tsx"
    tests: "co-located as {Name}.test.tsx"

preferences:
  state_management: "zustand"
  styling: "tailwind"
  avoid:
    - pattern: "class components"
      reason: "Use functional components with hooks"

caution:
  - path: "src/lib/payments/*"
    reason: "Payment processing - PCI compliance required"
    severity: "critical"

testing:
  framework: "vitest"
  run_command: "npm test"

history:
  created: "2024-01-15"
  last_updated: "2024-01-21"

Schema

The plugin includes a comprehensive JSON Schema (schemas/ai-context.schema.json) that defines all supported fields:

• Project types: web-app, api, cli, library, mobile-app, desktop-app, monorepo, microservices
• Component statuses: stable, unstable, deprecated, experimental
• Severity levels: info, warning, critical
• Naming conventions: camelCase, snake_case, PascalCase, kebab-case

Auto-Loading

The plugin sets up automatic context loading via:

1. SessionStart hook in .claude/settings.json (primary)
2. CLAUDE.md reference (fallback for other AI tools)

This ensures AI assistants always have your project context.

Security & Trust

• No external dependencies - All functionality is self-contained
• No data collection - Your context stays local
• No network calls - Except for Claude API (handled by Claude Code)
• Open source - Full transparency

Requirements

• Claude Code 1.0 or later
• No external dependencies

Contributing

Contributions welcome! See CONTRIBUTING.md for guidelines.

License

MIT License - see LICENSE for details.

---

Built for the ans marketplace

---

Built by AI Native Systems™