Commands Reference

The AI Context System v6.0 uses 8 slash commands — pure prompts with no scripts or machinery.

Command Overview

Command	Purpose	When to Use
/init-context	Create context files	Once, at project setup
/save	Update STATUS.md	End of every session
/update-context-system	Update commands (v6.x only)	When updates available
/review-security	Security audit	Before deployment
/review-performance	Performance check	Before deployment
/review-accessibility	Accessibility audit	Before deployment
/review-seo	SEO review	Before deployment
/review-cost	Cost optimization	When optimizing spend

Core Commands

/init-context

Creates the three context files for a new project:

• CLAUDE.md — Entry point (at project root)
• context/STATUS.md — Current state
• context/DECISIONS.md — Decision log

Usage:

/init-context

When: Once per project, at setup.

/save

Updates STATUS.md with current state. The core of the Session Loop.

Updates:

• LastUpdated date
• HeadCommit (git SHA)
• Objective (if changed)
• Working Set
• Next Actions
• Blocked On

Also: Autonomously evaluates session and records significant decisions to DECISIONS.md.

Usage:

/save

When: End of every session. Build the habit.

/update-context-system

Updates command files from v6.x to v6.y. For pre-v6 migrations, use the migrate-to-v6.sh script instead.

Usage:

/update-context-system

When: Periodically, when updates are available (v6.x projects only).

Review Commands

These are optional audits you can run before deployment. Each produces a report — no automatic changes.

/review-security

OWASP Top 10 security audit. Checks:

• Authentication and authorization
• Input validation and injection
• XSS and CSRF protection
• Secrets and credentials
• Dependency vulnerabilities

/review-performance

Performance review. Checks:

• Core Web Vitals (LCP, INP, CLS)
• Bundle size and code splitting
• Caching strategies
• Database query efficiency

/review-accessibility

WCAG 2.1 AA compliance audit. Checks:

• Keyboard navigation
• Screen reader compatibility
• Color contrast
• ARIA usage
• Focus management

/review-seo

Technical SEO review. Checks:

• Meta tags and OpenGraph
• Structured data (JSON-LD)
• Sitemap and robots.txt
• URL structure
• Canonical URLs

/review-cost

Cost optimization review. Serverless-first — checks:

• Rendering strategy (SSR vs ISR/SSG)
• Function duration and memory
• N+1 queries and connection pooling
• Image optimization and bandwidth
• Build minutes and cold starts
• API usage and caching

Command Philosophy

v6.0 commands are:

1. Pure prompts — No embedded bash scripts
2. Advisory — Claude follows guidelines, no mechanical validation
3. Simple — Each command does one thing well
4. Human-readable — Anyone can understand what a command does

Architecture

.claude/
├── commands/           # 8 slash command files
│   ├── init-context.md
│   ├── save.md
│   ├── update-context-system.md
│   ├── review-security.md
│   ├── review-performance.md
│   ├── review-accessibility.md
│   ├── review-seo.md
│   └── review-cost.md
└── VERSION             # Current version (6.0.3)

No agents. No schemas. No hooks. No scripts.

The Session Loop

The most important pattern:

Start → Read STATUS.md
Work → Edit Working Set files
End → Run /save

Follow this loop consistently and context persists naturally.