Openclaw Command Development

Command Development for OpenClaw

Overview

OpenClaw skills can include slash commands defined as Markdown files. These commands provide reusable, standardized prompts that agents execute during interactive sessions.

Key concepts:

• Markdown file format with optional YAML frontmatter
• Dynamic arguments and file references
• Skill-based organization in ~/.openclaw/skills/
• Commands are instructions FOR agents, not messages TO users

Command Basics

What is a Slash Command?

A slash command is a Markdown file containing a prompt that an OpenClaw agent executes when invoked. Commands provide:

• Reusability: Define once, use repeatedly
• Consistency: Standardize common workflows
• Sharing: Distribute across the agent roster
• Efficiency: Quick access to complex multi-step prompts

Critical: Commands are Instructions FOR Agents

When a command is invoked, its content becomes the agent's instructions. Write commands as directives TO the agent:

Correct (instructions for agent):

Review this code for security vulnerabilities including:
- SQL injection
- XSS attacks
- Authentication issues

Provide specific line numbers and severity ratings.

Incorrect (messages to user):

This command will review your code for security issues.

Command Locations

Personal commands (available everywhere):

• Location: ~/.openclaw/skills/<skill-name>/commands/
• Scope: Available in all agent sessions

Workspace commands (shared with all agents):

• Location: ~/.openclaw/workspace/commands/
• Scope: Available to all 16 agents in the roster

File Format

Basic Structure

~/.openclaw/skills/my-skill/
└── commands/
    ├── review.md           # /review command
    ├── test.md             # /test command
    └── deploy.md           # /deploy command

Simple command (no frontmatter needed):

Review this code for security vulnerabilities including:
- SQL injection
- XSS attacks
- Authentication bypass

With YAML Frontmatter

---
description: Review code for security issues
allowed-tools: Read, Grep, Bash(git:*)
---

Review this code for security vulnerabilities...

Frontmatter Fields

description

Brief description shown in help output.

description: Review pull request for code quality

allowed-tools

Specify which tools the command can use.

allowed-tools: Read, Write, Edit, Bash(git:*)

argument-hint

Document expected arguments for autocomplete.

argument-hint: [file-path] [priority]

Dynamic Arguments

Using $ARGUMENTS

Capture all arguments as a single string:

---
description: Fix issue by number
argument-hint: [issue-number]
---

Fix issue #$ARGUMENTS following our coding standards.

Positional Arguments ($1, $2, $3)

---
description: Review PR with priority
argument-hint: [pr-number] [priority]
---

Review pull request #$1 with priority level $2.

File References

Using @ Syntax

Include file contents in command:

---
description: Review specific file
argument-hint: [file-path]
---

Review @$1 for code quality and best practices.

Integration with OpenClaw Agents

Dispatching Commands to Specific Agents

Commands can be routed to specific agents in the 16-agent roster:

---
description: Deep security audit
---

Route this to Vault (cybersecurity agent) for analysis.

openclaw agent spawn --agent cybersecurity --task "Perform deep security audit on recent changes"

Multi-Agent Workflow Commands

---
description: Full review pipeline
argument-hint: [file-path]
---

Phase 1: Route to Sage (solution-architect) for design review
Phase 2: Route to Vault (cybersecurity) for security analysis
Phase 3: Route to Vex (tester) for test coverage assessment

Target: @$1

Command Organization

Flat Structure

commands/
├── build.md
├── test.md
├── deploy.md
└── review.md

Namespaced Structure

commands/
├── ci/
│   ├── build.md
│   └── test.md
├── git/
│   ├── commit.md
│   └── pr.md
└── docs/
    └── generate.md

Best Practices

1. Single responsibility: One command, one task
2. Clear descriptions: Self-explanatory
3. Document arguments: Always provide argument-hint
4. Consistent naming: Use verb-noun pattern (review-pr, fix-issue)
5. Agent-aware: Specify which agent(s) should handle when relevant
6. Workspace paths: Use ~/.openclaw/ paths, never hardcoded absolute paths

Common Patterns

Review Pattern

---
description: Review code changes
allowed-tools: Read, Bash(git:*)
---

Review each changed file for:
1. Code quality and style
2. Potential bugs
3. Test coverage
4. Documentation needs

Agent Dispatch Pattern

---
description: Spawn analysis task
argument-hint: [task-description]
---

openclaw agent spawn --task "$ARGUMENTS"

Additional Resources

Reference Files

• references/frontmatter-reference.md — Complete frontmatter field specifications
• references/advanced-workflows.md — Multi-step and multi-agent command patterns

Example Files

• examples/simple-commands.md — Basic command examples
• examples/plugin-commands.md — Skill-integrated command patterns