Agent Skills: OpenCode Commands

Use this skill when creating, modifying, or understanding OpenCode commands stored in `~/.config/opencode/command/`.

UncategorizedID: alexismanuel/dotfiles/opencode-command

Install this agent skill to your local

pnpm dlx add-skill https://github.com/alexismanuel/dotfiles/tree/HEAD/.config/opencode/skills/opencode-command

Skill Files

Browse the full folder contents for opencode-command.

Download Skill

Loading file tree…

.config/opencode/skills/opencode-command/SKILL.md

Skill Metadata

Name
opencode-command
Description
Use this skill when creating, modifying, or understanding OpenCode commands stored in `~/.config/opencode/command/`.

OpenCode Commands

Use this skill when creating, modifying, or understanding OpenCode commands stored in ~/.config/opencode/command/.

Location

Commands directory: ~/.config/opencode/command/ Vault root: /Users/alexismanuel/.config/opencode/

Command Structure

Commands are stored user prompts - markdown files with YAML frontmatter.

File Format

---
description: Short description of what command does
---

[Instructions for agent - what to do when command is invoked]

Load relevant skills if needed.

Arguments: $ARGUMENTS

Required Fields

  • description (YAML frontmatter): Short, action-oriented summary
  • $ARGUMENTS: Placeholder for command arguments (if needed)

Naming

  • Use kebab-case: bootstrap-daily-note.md
  • Descriptive, concise names matching the command's purpose

Creating Commands

  1. Choose a descriptive name (kebab-case)
  2. Create markdown file in ~/.config/opencode/command/
  3. Add YAML frontmatter with description
  4. Write clear instructions for the agent
  5. Reference relevant skills to load
  6. Include $ARGUMENTS placeholder if command accepts arguments

Examples

Existing commands to reference:

  • diary.md - Update daily note with session summary
  • save-session.md - Save OpenCode session to vault
  • bootstrap-daily-note.md - Bootstrap daily note from yesterday

Best Practices

  • Keep instructions concise and actionable
  • Always reference relevant skills
  • Test commands after creation
  • Use consistent formatting across commands