Skill Setup

msgraph follows the Agent Skills specification. This page explains how to install it for different AI agents.

Directory Structure

The skill is contained in the skills/msgraph/ directory of the repository. When installed, the msgraph/ directory is placed in your agent’s skill folder:

msgraph/
├── SKILL.md                          # Skill definition (required)
├── scripts/
│   ├── run.sh                        # macOS/Linux launcher
│   ├── run.ps1                       # Windows launcher
│   └── bin/                          # Binary cache (auto-populated)
└── references/
    ├── REFERENCE.md                  # API reference for agents
    ├── graph-api-index.db            # Pre-processed OpenAPI index (SQLite FTS5)
    ├── samples-index.db              # Samples index (SQLite FTS5)
    └── api-docs-index.db             # API docs index (SQLite FTS5)

No monorepo files (docs, CI, source code) are included in the installed skill.

Installation

Option 1: Using skills.sh (Recommended)

npx skills add merill/msgraph

This automatically discovers the skill and installs it into your agent’s skill directory.

Option 2: Download from GitHub Releases

Download msgraph.zip from the latest release and extract it:

curl -fsSL -o msgraph.zip https://github.com/merill/msgraph/releases/latest/download/msgraph.zip
unzip msgraph.zip -d ~/.agent/skills/

Option 3: Manual Symlink (for contributors)

If you’ve cloned the repo for development:

ln -s /path/to/msgraph/skills/msgraph ~/.agent/skills/msgraph

How the Agent Uses It

1. Discovery: The agent reads the SKILL.md frontmatter (name + description) at startup
2. Activation: When the user asks about Microsoft 365 data, the agent loads the full SKILL.md
3. Execution: The agent runs CLI commands via the launcher scripts
4. Reference: If needed, the agent reads references/REFERENCE.md for API details

Launcher Scripts

The launcher scripts (run.sh / run.ps1) handle platform management:

• Detect the platform (OS + architecture)
• Download the correct build for your platform
• Cache locally in scripts/bin/
• Forward all arguments

# Via launcher script
./msgraph/scripts/run.sh auth status

# Or directly with the CLI
msgraph auth status

Custom Client ID

If your organization requires a specific Entra ID app registration, set the environment variable before the agent runs:

export MSGRAPH_CLIENT_ID="your-custom-app-id"

The agent skill automatically picks this up.