Setup & Installation
Or with the ClawHub CLI, for registry-managed skill folders outside a full OpenClaw workspace:
What This Skill Does
Connects to Gmail and GitHub through a local access control gateway that the data owner configures. Agents can pull filtered, redacted data and propose outbound actions, which are staged for owner approval before anything executes. Every pull and propose is logged with an auditable purpose string.
Unlike direct API access, every agent action goes through an approval queue and policy filter, so the agent cannot read fields or send messages the owner hasn't explicitly allowed.
When to use it
- Summarizing unread emails from a specific sender
- Drafting email replies for review before sending
- Querying open GitHub issues assigned to you
- Triaging Q4 report threads without exposing full mailbox contents
- Maintaining an audit trail of what an agent read and why
Example Workflow
Here's how your AI assistant might use this skill in practice.
User asks: Draft a reply to Alice's Q4 report email thanking her for the numbers
- 1Calls personal_data_pull with source=gmail, query='from:alice subject:Q4 report', purpose='find email to reply to'
- 2Retrieves filtered email metadata and body per owner policy
- 3Calls personal_data_propose with action_type=reply_email, recipient, subject, and body populated
- 4Action is placed in the staging queue, not sent
- 5Owner reviews and approves the draft at http://localhost:3000
Draft reply queued for owner approval; nothing is sent until the owner confirms
Requirements
Accounts, API keys, or tools you or your AI assistant may need to set up while using this skill.