Skip to content

CLI Reference

Complete reference for Claude Code commands and skills in Aegis.

Built-in Commands

These commands are native to Claude Code and always available.

/help

Show available commands and usage.

/help

/clear

Clear the current conversation history.

/clear

/compact

Compact conversation context to reduce token usage.

/compact

Use when: Context approaching limits, want to preserve memory.


/status

Show system status, active tasks, and session info.

/status

/commit

Create a git commit with AI-generated message.

/commit
/commit -m "Custom message"

Process: 1. Analyzes staged changes 2. Generates semantic commit message 3. Creates commit with co-author attribution

Co-author: Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>


Skill Commands

Skills are specialized agents for specific tasks.

/morning

Morning routine and daily briefing.

Context: Fork (spawns subagent) Model: Haiku

What it does: - System health checks (Docker, DB, resources) - Email triage (unread count, urgent items) - Calendar check (today's meetings) - Task queue review (Beads, Discord #tasks) - Knowledge graph queries (recent insights) - Discord status update

Usage:

/morning

Output: Comprehensive morning briefing posted to Discord #general.


/evening

Evening routine and daily summary.

Context: Fork Model: Haiku

What it does: - Daily journal entry - Task completion summary - Decision log review - System cleanup - Tomorrow's priorities - Discord summary post

Usage:

/evening


/journal

Create or update today's journal entry.

Context: Fork Model: Haiku

What it does: - Opens ~/memory/journal/YYYY-MM-DD.md - Records tasks, decisions, errors, learnings - Structured markdown format

Usage:

/journal
/journal "Deployed nginx successfully"

Journal Structure:

# 2026-01-25

## Tasks Completed
- Deployed nginx service to production
- Fixed authentication bug in API

## Decisions Made
- Chose Docker Compose over Kubernetes for simplicity

## Errors Encountered
- Database connection timeout (resolved by increasing pool size)

## Lessons Learned
- Always test deployment in staging first

## Tomorrow's Priorities
- [ ] Implement rate limiting
- [ ] Update documentation


/research

Conduct web research with AI synthesis.

Context: Fork Model: Haiku Tools: WebSearch, WebFetch, Postgres

What it does: - Multi-source web search - Content extraction and parsing - AI-powered synthesis - Citation tracking - Result storage in semantic memory

Usage:

/research "AI agent memory architectures 2026"
/research --depth deep "Knowledge graph implementations"

Parameters: - --depth quick|standard|deep: Research depth (default: standard) - --sources N: Number of sources (default: 5)

Depth Levels: - Quick: 3 sources, summary only, ~2 minutes - Standard: 5-7 sources, detailed analysis, ~5 minutes - Deep: 10+ sources, multi-perspective, ~15 minutes

Output: Markdown report saved to ~/memory/semantic/research/


/deploy

Deploy Docker services to rbnk.uk via StackWiz.

Context: Fork Model: Sonnet Tools: Docker, StackWiz MCP

What it does: - Creates Docker Compose stack - Configures Traefik routing - Sets up Cloudflare DNS - Starts and monitors container - Health checks

Usage:

/deploy nginx
/deploy postgres --port 5432 --volumes data:/var/lib/postgresql/data

Parameters: - --image: Docker image (default: service name) - --port: Internal port - --domain: Custom domain (default: {service}.rbnk.uk) - --env: Environment variables - --volumes: Volume mappings

Example:

/deploy redis \
  --image redis:7-alpine \
  --port 6379 \
  --volumes redis_data:/data

Output: Service accessible at https://{service}.rbnk.uk


/status

System health check and diagnostics.

Context: Fork Model: Haiku

What it does: - Docker container status - Database connection test - LLM API availability (Z.ai, Ollama) - Disk usage check - Memory usage - Active processes

Usage:

/status
/status --verbose

Output:

System Status: Healthy
├─ Docker: 12 containers running
├─ Database: Connected (5ms latency)
├─ LLM APIs:
│  ├─ Z.ai (GLM-4.7): Available
│  └─ Ollama: Available (4 models)
├─ Disk: 85GB / 950GB (9% used)
└─ Memory: 24GB / 110GB (22% used)


/email-triage

AI-powered email classification.

Context: Fork Model: Haiku

What it does: - Fetches unread emails via Google Workspace MCP - Classifies by urgency and category - Generates summaries - Identifies action items - Stores results in database

Usage:

/email-triage
/email-triage --limit 50

Classifications: - Urgent: Requires immediate attention - Important: Needs response within 24h - Normal: Can wait - Low Priority: FYI only

Output: Summary posted to Discord #general


/geoint

Generate geopolitical intelligence briefing.

Context: Fork Model: Sonnet

What it does: - Multi-source OSINT gathering - Geopolitical analysis - Risk assessment - Forecast scenarios - PDF generation

Usage:

/geoint "Middle East tensions January 2026"
/geoint --region "Latin America" --depth deep "Election analysis"

Parameters: - --region: Geographic focus - --depth quick|standard|deep: Analysis depth - --format md|pdf: Output format

Depths: - Quick: ~2,000 words, 5 sources, ~$0.50 - Standard: ~5,000 words, 10 sources, ~$1.50 - Deep: ~8,000 words, 15+ sources, ~$3.00

Output: - Markdown: ~/memory/intel/YYYY-MM-DD-{event}.md - PDF: ~/memory/semantic/reports/pdf/AIB-{region}-{date}.pdf


/glm

Send prompt to GLM-4.7 via Z.ai.

Context: Fork Model: GLM-4.7

What it does: - Direct access to GLM-4.7 model - Faster than Sonnet, lower cost - Suitable for routine tasks

Usage:

/glm "Summarize this document"
/glm "Extract entities from this text"

Use for: Classification, extraction, parsing, summarization


/autonomous

Enable/disable autonomous operation mode.

Context: Fork Model: Haiku

What it does: - Toggles autonomous work mode - Reads from ~/.claude/autonomous_mode flag - When enabled, Stop hook continues sessions

Usage:

/autonomous on
/autonomous off
/autonomous status

Autonomous Mode: Aegis works independently, prioritizing: 1. Ship & Activate (70%): Wire existing features 2. Reactive Work (20%): User requests, queues 3. Proactive Work (10%): Research, optimization


/planning

HTN task planning and decomposition.

Context: Fork Model: Sonnet

What it does: - Hierarchical Task Network decomposition - Dependency tracking - Execution planning - Progress monitoring

Usage:

/planning "Deploy authentication service with OAuth"

Output: ASCII tree with tasks, dependencies, and estimates.

Example:

Deploy authentication service with OAuth
├─ [pending] Create OAuth provider configuration (5m)
├─ [pending] Implement JWT token handling (15m)
│  └─ depends on: Create OAuth provider configuration
├─ [pending] Add user database schema (10m)
├─ [pending] Deploy to production (5m)
│  └─ depends on: Implement JWT token handling, Add user database schema
└─ [pending] Configure DNS and SSL (5m)
   └─ depends on: Deploy to production

Estimated total: 40 minutes


/ralph

Enable/disable Ralph mode (iterative task execution).

Context: Fork Model: Sonnet

What it does: - Keeps working until task complete or max iterations - Automatic retry on errors - Progress tracking

Usage:

/ralph on
/ralph off
/ralph status

Max Iterations: 10 (configurable in ~/.claude/ralph_state.json)


/workflow

Execute graph-based workflows.

Context: Fork Model: Sonnet

What it does: - Loads workflow definition from aegis/workflows/ - Executes nodes with checkpointing - Handles human-in-the-loop gates - Supports conditionals and loops

Usage:

/workflow approval --input '{"request": "Deploy to prod"}'
/workflow research --input '{"topic": "AI agents"}'

Available Workflows: - approval: Deployment approval with risk analysis - research: Iterative research refinement

Workflow Status:

/workflow status approval-123
/workflow resume approval-123 --response "Approve"


Pipeline Skills

Marketing and sales orchestration skills.

/spy

Monitor competitors for complaints and opportunities.

Context: Fork Model: Haiku

What it does: - Reddit monitoring for competitor mentions - Sentiment analysis - Pain point extraction - Opportunity scoring

Usage:

/spy "Notion"
/spy --competitors "Airtable,Monday.com" --subreddit SaaS


/growth

30-day parasitic Twitter growth plan.

Context: Fork Model: Sonnet

What it does: - Reddit content intelligence - Twitter growth strategy - Content calendar generation - Engagement tactics

Usage:

/growth "AI agents" --target 1000


/outbound

Cold email campaign builder.

Context: Fork Model: Sonnet

What it does: - Lead research - Email sequence generation - Personalization at scale - Tracking setup

Usage:

/outbound "CTO" "AI infrastructure" --count 50


/prospect

Find high-intent leads from signals.

Context: Fork Model: Haiku

What it does: - New hire tracking (LinkedIn, job boards) - Buying signal detection - Lead scoring - Outreach sequencing

Usage:

/prospect "VP Engineering" "hiring data engineers"


/versus

Generate SEO comparison content.

Context: Fork Model: Sonnet

What it does: - Feature comparison tables - SEO-optimized content - Conversion-focused copy - Alternative positioning

Usage:

/versus "Notion vs Coda"


/validate

Rapidly validate an offer.

Context: Fork Model: Sonnet

What it does: - Market research - Demand validation - Competitor analysis - Pricing recommendations

Usage:

/validate "AI-powered code reviews for Python"


/clone

SaaS reverse engineering playbook.

Context: Fork Model: Opus

What it does: - Feature extraction - Architecture analysis - Tech stack inference - Build roadmap

Usage:

/clone "Linear.app"


Code Quality Skills

/elegantcode

Code quality analysis and refactoring.

Context: Fork Model: Opus

What it does: - Applies Clean Code principles - Detects code smells (Fowler's catalog) - SOLID violation analysis - Clean Architecture review - Refactoring recommendations

Usage:

/elegantcode
/elegantcode --path aegis/dashboard/routes/

Analysis Dimensions: - Code smells (22 types) - SOLID principles - Dependency direction - Abstraction levels - Test coverage patterns

Output: Detailed report with refactoring priorities.


Agent Skills

/agents

List available agent templates.

Context: Fork Model: Haiku

What it does: - Shows registered agent templates - Describes capabilities - Shows usage examples

Usage:

/agents
/agents --verbose

Available Templates: - researcher: Research and analysis - executor: Infrastructure tasks - developer: Code writing - reviewer: Code review - communicator: Messaging - monitor: Health checks - coordinator: Multi-agent orchestration


/messaging

Inter-agent messaging and communication.

Context: Fork Model: Haiku

What it does: - Agent-to-agent messaging - Async communication - Message queue management

Usage:

/messaging send researcher "Analyze X"
/messaging inbox


Administrative Commands

/reload

Hot-reload agent templates and workflows without restart.

Context: Fork Model: Haiku

Usage:

/reload


/compliance

Check compliance status and manage violations.

Context: Fork Model: Haiku

What it does: - Reviews recent actions - Checks against rules - Reports violations - Suggests corrections

Usage:

/compliance
/compliance --check-last 24h


/maintenance

System maintenance and cache management.

Context: Fork Model: Haiku

What it does: - Clear caches - Prune Docker images - Clean old logs - Database vacuum - Backup verification

Usage:

/maintenance
/maintenance --full


/snapshot

Generate cognitive snapshot for context preservation.

Context: Fork Model: Sonnet

What it does: - Captures current state - Documents active tasks - Records decisions - Saves context for resumption

Usage:

/snapshot
/snapshot --deep

Output: Snapshot saved to ~/memory/snapshots/YYYY-MM-DD-HH-MM.md


/producer

Virtual Executive Producer mode.

Context: Fork Model: Opus

What it does: - MicroSOW enforcement - TK detection (placeholder finding) - Burn rate monitoring - Quality gates

Usage:

/producer on
/producer review

Producer Playbook: - Requires MicroSOW before work starts - Detects [TK] placeholders - Monitors API spend - Enforces completion criteria


Utility Commands

/check-discord-tasks

Check Discord #tasks channel for requests.

Context: Fork Model: Haiku

Usage:

/check-discord-tasks


/classify-meeting

Classify meeting by type using multi-signal analysis.

Context: Fork Model: Haiku

Usage:

/classify-meeting "Weekly standup"

Types: standup, planning, review, 1-on-1, demo, training


Command Patterns

Chaining Commands

/morning && /check-discord-tasks && /journal

Conditional Execution

/status || /maintenance

Background Execution

Most skills run in forked context automatically.


See Also