Commands are single-shot, named tasks that agents or users can invoke directly. Unlike skills (which describe how to do something), commands describe what to do in a specific, repeatable way.
Think of them as slash commands, macros, or saved prompts.
.agents/commands/
├── explain-error.md
├── summarize-diff.md
└── check-conventions.md
---
name: explain-error
description: Explain a code error in plain language. Use when the user pastes an error message.
input: error_message
---
## Task
Given the error message provided, explain:
1. What caused the error (in plain language)
2. Where in the code it likely originated
3. How to fix it (step by step)
4. How to prevent it in the future
## Output format
- Use simple language, avoid jargon
- Max 200 words
- Code examples where relevant| Field | Required | Description |
|---|---|---|
name |
Yes | Unique identifier, kebab-case |
description |
Yes | When to use this command |
input |
No | What input the command expects |
Commands are Tier 1 for names/descriptions, Tier 2 for full content on invocation. Users can invoke commands explicitly (e.g., /explain-error) or agents can activate them automatically when the task matches.