OpenCode custom commands: .opencode/commands, $ARGUMENTS, shell output and subtasks

• Per project: .opencode/commands/NAME.md.
• Global: ~/.config/opencode/commands/NAME.md.
• In config: the "command" option of your OpenCode config, where each entry needs a template.
• A custom command with the same name as a built-in one, such as /init, /undo, /redo or /help, overrides it.

---
description: Run tests with coverage
agent: build
---
Run the full test suite with coverage for $ARGUMENTS.

Recent commits:
!`git log --oneline -10`

Focus on failing tests and suggest fixes.

Run it with /test or /test src/auth. The file name sets the command name; the frontmatter is optional.

• template: the prompt sent to the model. Required in JSON config; in a Markdown file the body is the template.
• description: shown in the TUI when you type the command.
• agent: which agent runs the command. If it is a subagent, the command runs as a subagent invocation by default.
• subtask: true forces a subagent invocation so the command stays out of your primary context; false turns the default subagent behavior off.
• model: overrides the default model for this command, for example anthropic/claude-3-5-sonnet-20241022.

• $ARGUMENTS is replaced with everything you type after the command name.
• $1, $2, $3 and so on give you individual arguments.
• !`command` inserts the output of a shell command. Commands run in the project root.
• @path includes the content of a file, for example @src/components/Button.tsx.

{
  "$schema": "https://opencode.ai/config.json",
  "command": {
    "component": {
      "template": "Create a new React component named $ARGUMENTS with TypeScript support.",
      "description": "Create a new component",
      "agent": "build",
      "subtask": false
    }
  }
}

• Compare: custom slash commands in every AI coding agent
• OpenCode config (opencode.json)
• OpenCode agents and subagents
• OpenCode permissions