A secure Gmail proxy for AI agents that enforces compliance through policies, audit logging, and human-in-the-loop approvals.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install clawback-gmail
2. Manual Installation
Copy the skill folder to one of these locations
~/.openclaw/skills/ <project>/skills/ Priority: Workspace > Local > Bundled
3. Prompt Installation
Copy this prompt to OpenClaw to install it automatically.
Help me install clawback-gmail using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).
Prefer to download?
Get the raw skill files in a ZIP archive.
What is clawback?
clawback is a specialized security proxy designed to provide a secure interface for Gmail operations. It acts as a gatekeeper between the user (or AI agent) and the Gmail API, ensuring that every action—whether reading, searching, or sending emails—complies with predefined organizational policies. By routing all traffic through a server-side proxy, clawback provides a robust audit trail and prevents unauthorized data exfiltration.
This utility is particularly powerful for environments where AI agents handle communication. It introduces a human-in-the-loop mechanism where sensitive actions, such as sending emails to external domains, can be intercepted and held for manual approval. This ensures that while agents are productive, they remain within safe operational boundaries when using Openclaw Skills.
clawback Use Cases
- Enforcing organizational security policies on all outbound emails sent by AI agents.
- Auditing Gmail interactions to maintain a clear record of communication history and compliance.
- Implementing approval workflows for sensitive actions like deleting messages or sending to specific recipients.
- Searching and managing large volumes of email threads with automated pagination and JSON output.
- Securely managing Gmail labels, drafts, and attachments through a standardized CLI interface.
How clawback Works
- The user or agent initiates a Gmail command via the clawback CLI tool.
- The request is routed to the server-side proxy which evaluates the action against active security policies.
- If a policy requires human intervention (e.g., approve_before_send), the command pauses and returns exit code 8.
- A human reviewer approves or rejects the action via a dashboard or email link.
- Upon approval, the proxy executes the requested Gmail operation and logs the transaction for audit purposes.
clawback Setup
First, install the tool using the official installation script:
curl -fsSL https://clawback.sh/install | bash
Authenticate with your Google account via the device flow:
clawback auth login
Verify your connection status to ensure the proxy is reachable:
clawback auth status
(Optional) If using a private proxy instance, configure the server environment variable:
export CB_SERVER=https://your-server.example.com
clawback Data Schema & Taxonomy
clawback utilizes a structured approach to Gmail data, focusing on JSON output for seamless integration with other Openclaw Skills.
| Entity | Data Type | Description |
|---|---|---|
| Message | JSON | Contains messageId, threadId, headers, and body content. |
| Thread | JSON | Grouped messages with threadId and label information. |
| Approval | JSON | Record of a pending action including approvalId and status. |
| Policy | JSON | Defined rules governing read/write permissions. |
| History | JSON | Chronological log of changes to the mailbox since a specific ID. |
Common flags for data manipulation include --select to project specific fields and --results-only to return raw data arrays without the envelope.
clawback Advanced Features
- Policy-driven Interception: Automatically blocks or flags actions based on granular, server-side rules.
- Auto-pagination: Use the --all flag to automatically fetch all results for large search queries or thread lists.
- Batch Operations: Perform actions like batch delete or batch modify across multiple message IDs in a single command.
- Multi-connection Support: Manage multiple Gmail accounts using the --connection flag to specify which account to use.
- Scripting Optimization: Headless mode support using --json, --no-input, and --fail-empty for robust automated pipelines.
SKILL.md
Loading
Related Openclaw Skills
Topic to Article Kit
A specialized research automation tool that converts AI topics into comprehensive writing packages including social evidence and structured outlines in Obsidian.
Creator Alpha Feed
A specialized automation tool for scouting, ranking, and archiving daily AI industry insights and KOL activity into structured Obsidian reports.
OpenRouter Usage Monitor
A dedicated tool for monitoring real-time OpenRouter API spend, model activity, and account balances directly within your agent workflow.
Agent Identity
A cryptographic identity framework for AI agents to prove their identity, verify others, and build on-chain reputation using USDC staking.
OpenClaw Update Skill
A specialized management tool for checking, applying, and verifying updates to the OpenClaw platform and its core components.
OpenClaw Update
A dedicated management skill for monitoring and applying version updates to the OpenClaw platform.