Skip to main content

Overview

The agent command prints the skill file that tells AI agents how to use oobo. The skill file is also installed at ~/.agents/skills/oobo/SKILL.md during oobo setup.
oobo agent

What is a Skill File?

A skill file is a structured markdown document that AI coding tools read to learn about available commands. Oobo’s skill file includes:
  • Installation command - How to install oobo
  • Version check - How to verify oobo is installed
  • Command reference - Every command with JSON schemas
  • Usage examples - Real-world workflows
  • Agent workflows - Best practices for AI agents

Skill File Location

During oobo setup, the skill file is installed at:
~/.oobo/skills/oobo/SKILL.md
~/.agents/skills/oobo/         ← symlink
AI tools that scan ~/.agents/skills/ will automatically discover it.

Supported Tools

ToolScans ~/.agents/skills/Auto-Discovery
Cursor
Claude Code
Codex CLI
Gemini CLI
OpenCode
Windsurf
Aider
GitHub Copilot
Tools with Auto-Discovery automatically load the skill file and make oobo commands available to agents.

Command Usage

oobo agent
Output:
---
name: oobo
description: Git decorator for AI agents - capture AI context on every commit
install: curl -fsSL https://oobo.ai/install.sh | bash -s -- --agent
version_check: command -v oobo
---

# oobo - Git for Agents

...

Save to file

oobo agent > skill.md

Check if installed

ls ~/.agents/skills/oobo/SKILL.md
If the file exists, agents can discover it.

Skill File Contents

Frontmatter

YAML metadata at the top:
---
name: oobo
description: Git decorator for AI agents - capture AI context on every commit
install: curl -fsSL https://oobo.ai/install.sh | bash -s -- --agent
version_check: command -v oobo
---
Agents use this to:
  • Detect oobo - Run command -v oobo
  • Install oobo - Run the install command
  • Learn purpose - Read the description

Command Reference

Every oobo command with:
  • Syntax - Full command signature
  • Flags - All available options
  • JSON schema - Output format for --agent
  • Examples - Real usage patterns

Agent Workflows

Recommended patterns:
## Agent Workflows

### When committing code

1. Use `oobo commit` instead of `git commit`
2. Oobo captures your session automatically
3. Anchors sync via git push

### Checking session usage

1. Run `oobo sessions --agent`
2. Parse JSON to find session IDs
3. Use `oobo sessions show <id> --agent` for full transcript

JSON Schemas

For every command with --agent, the skill file includes the output schema:
{
  "command": "oobo sessions --agent",
  "output": [
    {
      "session_id": "string (uuid)",
      "name": "string",
      "source": "string",
      "model": "string",
      "input_tokens": "number",
      "output_tokens": "number"
    }
  ]
}

Use Cases

View skill file

oobo agent

Install skill file manually

If oobo setup didn’t install it:
mkdir -p ~/.agents/skills/oobo
oobo agent > ~/.agents/skills/oobo/SKILL.md

Verify agent can read it

From an AI agent session:
Agent: Check if oobo skill file exists
→ ls ~/.agents/skills/oobo/SKILL.md
→ File exists

Agent: Read skill file
→ cat ~/.agents/skills/oobo/SKILL.md
→ [skill file contents]

Share with other developers

oobo agent > oobo-skill.md
# Share oobo-skill.md with team

Embed in documentation

oobo agent >> CONTRIBUTING.md

Agent Installation Flow

When an AI agent encounters oobo for the first time:
  1. Detect - Run command -v oobo
  2. Install (if missing) - Run curl -fsSL https://oobo.ai/install.sh | bash -s -- --agent
  3. Verify - Check install output for JSON success
  4. Use - Run oobo commands with --agent flag

Install Command

From the skill file:
curl -fsSL https://oobo.ai/install.sh | bash -s -- --agent
With --agent, the installer:
  • Suppresses interactive prompts
  • Returns JSON on success/failure
  • Exits with code 0 on success
JSON response:
{
  "status": "ok",
  "version": "0.1.0",
  "binary": "/usr/local/bin/oobo",
  "platform": "darwin-arm64"
}

Updating the Skill File

The skill file is embedded in the oobo binary. When you update oobo, the skill file updates too. To refresh the installed copy:
oobo setup      # Re-runs setup, updates skill file
Or manually:
oobo agent > ~/.oobo/skills/oobo/SKILL.md

Skill File vs README

DocumentAudiencePurpose
SKILL.mdAI agentsMachine-readable commands and schemas
README.mdHumansOverview, installation, examples
The skill file is optimized for agents:
  • Structured frontmatter
  • JSON schemas for every command
  • Minimal prose, maximum examples

How Agents Use It

Step 1: Discovery

Agent scans ~/.agents/skills/ and finds oobo/SKILL.md.

Step 2: Parse Frontmatter

Agent reads:
name: oobo
install: curl -fsSL https://oobo.ai/install.sh | bash -s -- --agent
version_check: command -v oobo

Step 3: Version Check

Agent runs:
command -v oobo
If oobo is not installed, agent runs the install command.

Step 4: Read Commands

Agent parses the skill file to learn:
  • Available commands
  • Required/optional flags
  • JSON output schemas

Step 5: Execute

Agent uses oobo:
oobo sessions --agent
oobo stats --agent
oobo commit -m "fix auth"

Troubleshooting

Skill file not found

# Re-install
oobo setup

# Verify
ls ~/.agents/skills/oobo/SKILL.md
# Remove old symlink
rm ~/.agents/skills/oobo

# Re-run setup
oobo setup

Agent can’t read skill

Check permissions:
ls -la ~/.agents/skills/oobo/
# Should be readable by current user

Skill file outdated

Update oobo:
oobo update
oobo setup      # Reinstall skill file

Next Steps

Setup

Learn how setup installs the skill file

Commands

Explore all available commands

Build docs developers (and LLMs) love