Slash Commands

Slash commands are shortcuts that trigger specialized AI workflows. Type / in the chat input to see available commands, then select one from the popup.

How Slash Commands Work

When you type / in the chat input, a popup appears listing all available commands. Use Arrow keys to navigate and Enter or Tab to select.

Important: Always select the slash command from the popup menu. If you type the command manually without selecting it, Invoke treats it as a regular message and sends it to the default Agent — it won't trigger the specialized workflow.

When you select a command, a styled pill (e.g., /skill) appears in the input along with a placeholder prompt. Replace the placeholder with your actual request and press Enter to send.

Available Commands

/invoke-guide

Ask questions about Invoke's features, shortcuts, settings, agent modes, and workflows.

Examples:

/invoke-guide What keyboard shortcuts are available?
/invoke-guide How do I set up MCP servers?
/invoke-guide What's the difference between Agent and Plan mode?

/summary

Condense the current conversation into a compact summary card.

The summary includes key topics discussed, decisions made, code changes, and outstanding tasks. Useful when your conversation is getting long and approaching context limits.

How to use: Type /, select summary from the popup, and press Enter. No additional input needed.

/skill

Create a new reusable AI skill — a knowledge bundle with instructions, scripts, references, and assets.

How to use: Select /skill from the popup, then describe what the skill should do:

/skill Create a code review skill that checks for security vulnerabilities and follows OWASP guidelines
Scope: Project

The AI will create a skill folder with a SKILL.md file and any necessary scripts or references.

/agent

Create a new custom agent with its own system prompt and tool access.

How to use: Select /agent from the popup, then describe the agent:

/agent Create a documentation writer agent that generates clear API docs with examples. It should only read and write files, no terminal access.

The AI will create the agent in Settings → Agents with the appropriate configuration.

/subagent

Create a new subagent — a specialized agent that the main Agent can automatically delegate tasks to during execution.

How to use: Select /subagent from the popup, then describe what it should handle:

/subagent Create a test writer subagent that generates unit tests following the project's existing test patterns
Scope: Project

Unlike custom agents (which you switch to manually), subagents are called automatically by the main Agent when it determines a task needs specialized handling.

/boards

Manage your feature boards directly through the AI. Create boards, add tasks, move items, and organize your workflow.

How to use: Select /boards from the popup, then describe what you want:

/boards Create a new board called "Sprint 4" with columns: To Do, In Progress, Review, Done

/api

Scan the current project for API endpoints and add them to the Invoke API view.

How to use: Select /api from the popup, optionally describe which APIs to scan for:

/api Scan all REST endpoints in the Express server

Or leave the placeholder empty to scan the entire project. The AI reads your code, identifies endpoints, and adds them to the API view (Cmd+8) organized under base URLs with methods, headers, and request chaining.

/automation

Create and manage automated workflows that run manually or on a schedule.

How to use: Select /automation from the popup, then describe your workflow:

/automation Create a daily test report that runs npm test and saves results to TEST_REPORT.md every morning at 9 AM

The AI guides you through setting up steps, choosing a trigger (manual or scheduled), and managing your automation.

/exit

Exit a locked slash command session and return to the default agent.

How to use: Select /exit from the popup and press Enter.

Agent Command Locking

When you use /boards, /api, or /automation, the agent locks to that command's specialized tools. This means all follow-up prompts continue using the same toolset — you don't need to type the slash command again for each message.

For example, after sending /boards Create a sprint board, you can follow up with:

"Add a task called Fix login bug to the To Do column"
"Move it to In Progress"

These messages automatically go through the boards tools.

To unlock the agent and return to the default mode, use the /exit slash command.

Agent command locking only applies to /boards, /api, and /automation. Other slash commands like /skill, /agent, and /subagent do not lock the agent.

Creating Custom Slash Commands

You can create your own slash commands tailored to your workflow. Custom commands are stored as markdown files in your project.

How to Create a Custom Command

Create the commands folder

In your project root, create a .invoke/commands/ directory (if it doesn't exist).

Create a markdown file

Create a new markdown file with the command name. For example: .invoke/commands/review-code.md

The filename becomes the command name (without the extension). So review-code.md becomes /review-code.

Write the command prompt

Write clear instructions in markdown. This is what the AI will follow when you invoke the command:

# Code Review Command

You are an expert code reviewer. When the user provides code, analyze it for:

- **Performance issues** — Identify slow algorithms or inefficient patterns
- **Security vulnerabilities** — Check for XSS, injection, authentication issues
- **Code style** — Enforce consistency with the project's patterns
- **Readability** — Suggest clearer naming and structure
- **Testing** — Comment on test coverage

Provide feedback in this format:
- Critical issues (must fix)
- Suggestions (nice to have)
- Positive observations (done well)

Use the command

Type / in the chat and select your new command from the popup. It will appear with the standard command styling.

Command Naming Rules

Use lowercase letters, hyphens, and numbers
Keep it short and descriptive
No spaces or special characters
Examples: review-code, api-docs, test-generator

Custom commands are project-level only. They're stored in .invoke/commands/ and only available in that project. To share commands across projects, store them in ~/.invoke/commands/ globally.

Example: Creating a Test Writer Command

Create .invoke/commands/write-tests.md:

# Test Writer Command

You are a test automation expert. Write comprehensive tests following these rules:

1. **Test structure** — Use describe/it blocks with clear names
2. **Coverage** — Test happy path, edge cases, and error states
3. **Mocking** — Mock external dependencies and APIs
4. **Assertions** — Use specific assertions that fail with clear messages
5. **Setup/Teardown** — Include proper initialization and cleanup

Follow the project's existing test patterns and naming conventions.

When the user provides code or a file path, write tests for it.

Then use /write-tests anytime you need tests generated.

Tips

Use the popup — Always select from the / popup. Typing /skill manually sends it as plain text to the default agent.
Replace the placeholder — After selecting a command, a bracketed placeholder appears. Replace it with your specific request before sending.
Scope field — For /skill and /subagent, you can specify Scope: Project (current project only) or Scope: Global (all projects).
Iterate — You can use the same slash command multiple times in a conversation to create multiple skills, agents, or subagents.

On this page