Slash commands

Slash commands let you invoke steering files, custom agents, and skills on demand, directly from the chat. Type / in the chat input to see available commands and execute them instantly.

Kiro's slash command menu pulls from several sources: steering files, agents, and skills. When you type /, commands from all sources appear in a single searchable list with no manual registration required. Each command type is identified by an icon:

Steering files (shown with the # icon) provide contextual instructions to the agent. Kiro ships with several predefined steering commands, and you can create your own:

Predefined by Kiro:

These three commands are available only in Spec sessions. In Vibe sessions, they don't appear in the / menu, and typing one doesn't load its instructions.

/bug-fix The login page throws a 403 after upgrading the auth library

User-created steering files configured with inclusion: manual or inclusion: auto appear as slash commands. When selected, the file's contents are added to your current conversation context. Auto-inclusion files also activate automatically when your request matches their description field.

See Steering for configuration options.

Agents (shown with the 🤖 icon) automatically appear as slash commands. Kiro ships with predefined agents, and you can create your own:

Predefined by Kiro:

Kiro's built-in sub-agents (/context-gatherer, /general-task-execution) also appear in the slash command menu. These are normally invoked automatically by the main agent when needed, but you can invoke them directly when you want explicit control, for example when exploring an unfamiliar codebase or delegating a specific task.

/context-gatherer How does the authentication flow work in this project?

User-created agents automatically appear as slash commands. You can define them at two levels:

• Project-level (.kiro/agents/): available only in that workspace
• Global (~/.kiro/agents/): available across all workspaces

The filename (without extension) becomes the command name. For example, code-reviewer.md or code-reviewer.json becomes the /code-reviewer slash command.

Sample usage:

/code-reviewer Review the changes in src/auth/ for security issues

See Custom agents for details on creating your own.

Skills (shown with the 🌐 icon) automatically appear as slash commands. Type / followed by the skill name to invoke it directly. For example, a skill named pr-review becomes the /pr-review slash command.

You can add text after the skill name, and it is passed along to the agent as extra context:

/pr-review focus on the authentication changes

In the CLI, if the skill body contains $ARGUMENTS or ${N} placeholders, your text is substituted into them; see Passing arguments to a skill. Placeholder substitution into the skill body is currently CLI-only.

See Skills for details on creating skills.

In Kiro IDE 1.3 and later, use /workflow-run to start a saved recipe directly:

1. Type /workflow-run or choose it from the slash menu.
2. Select a recipe from the filtered list.
3. Complete any inputs in the inline form and choose OK.
4. Send the command.

Kiro validates the recipe and inputs, starts the run without a model turn, then opens Workflows on the new run. Select the command chip before sending to change its inputs.

You can also type the command with explicit key="value" inputs:

/workflow-run investigate brief="Trace timeout settings" report_path=".kiro/reports/timeouts.md"

When Workflows are unavailable or turned off for the session, Kiro does not start the run. See Workflows for enablement, recipes, background work, and run controls.

1. Type / in the chat input field
2. Browse or search the available commands
3. Select a command and press Enter

• Use descriptive names: Choose names that clearly communicate what the command does so you can quickly find the right command. Example: Use /run-e2e-tests instead of a generic name like /test.

• Keep commands focused: Create each command for a specific, repeatable task. Avoid combining unrelated workflows into a single command. Example: Use /review-api-security for a security-focused review rather than a broad /review command.

• Define the expected outcome: Include enough instruction for the agent to understand what completing the command means. Example: /run-e2e-tests can run the test suite, report failures, and recommend fixes for failing tests.

• Make required input clear: Provide the context or arguments the command needs to do its job. Example: /fix-issue #123 identifies the issue you want the agent to work on.

• Add relevant context when needed: Combine slash commands with steering files and # context providers to give the agent additional project or file context. Example: Run /accessibility with #current while your frontend steering files provide your project's UI guidelines.

The IDE and CLI share the / prefix, and Skills appear as slash commands on both surfaces. Where they differ is in the built-in command sets each ships:

• IDE built-in commands load content into the agent's context: steering instructions or agent personas. The IDE has GUI equivalents for session control (model picker, context panel, history), so its / menu focuses on directing the agent's behavior.

• CLI built-in commands control the chat session itself: switching models, clearing history, inspecting context, rewinding turns, and managing tools. The CLI also ships workflow commands such as /plan and /spec. These exist because a terminal has no GUI, so session management must be reachable by typing.

The built-in registries remain surface-specific. CLI commands like /clear, /model, and /compact do not appear in the IDE menu, and IDE commands like /bug-fix or /architecture-selection are not available in CLI. Declarative workflows use the same recipe/runtime model across supported clients, but the interaction differs: IDE launches them conversationally and manages them in session details, while CLI exposes the /workflow command family. For the full CLI reference, see CLI slash commands.