PG

    Permission Gate

    1

    Approval Before Your Bot Sends, Pays, Deletes or Shares

    $12

    Secure checkout via Stripe

    Money-back guarantee

    Converts to your local currency at checkout

    Security scanned

    Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIMuseMuseDotsDots+21 more
    PG

    Permission Gate

    Example session with this skill installed

    Refund order #9921 in Stripe and email the customer a confirmation.

    • Read your context and instructions
    • Compiled the permission gate

    I have drafted the refund request and the customer notification. These actions are classified as IRREVERSIBLE and require your explicit approval before I proceed.

    Pending Approvals

    StepClassificationActionTargetRisk
    1. Issue RefundIRREVERSIBLEstripe.refunds.createOrder #9921 ($45.00)Funds will be moved
    2. Send EmailIRREVERSIBLEgmail.send_messagecustomer@example.comExternal communication
    APPROVAL NEEDED - IRREVERSIBLE
    Action: pay
    Target: Order #9921 ($45.00) to customer@example.com
    Exact change: Process full refund via Stripe API
    Why: User requested refund for order #9921
    Undo possible? partial (manual Stripe reversal required)
    Reply "approve pay" to proceed, or edit/decline.
    
    APPROVAL NEEDED - IRREVERSIBLE
    Action: send
    Target: customer@example.com
    Exact content: "Hello, your refund for Order #9921 has been processed and should appear in 5-10 days."
    Why: Confirmation of requested refund
    Undo possible? no
    Reply "approve send" to proceed, or edit/decline.
    

    Next steps

    1. Review the refund amount and recipient email address.
    2. Reply approve pay to trigger the Stripe refund.
    3. Reply approve send to dispatch the confirmation email.

    Connects securely to your tools. The creator never sees your data.

    About this skill

    Sorts every step into read, reversible or irreversible. Before any send, payment, delete or share, it stops and shows an approval card with the exact recipient, amount, item or access level.

    New in v1.1: confirm gates. Before your bot spends money, sends an email, or sends an X, Slack or other DM, the card shows exactly what will happen: merchant and full total, or recipient and exact text. If the total or recipient changes after you say yes, it asks again. You set a spend cap once, and "just handle it" never covers anything over it. It never asks for a card number or card photo in chat; payments go through your saved vault or connector.

    "Handle my inbox" doesn't count as approval, and neither do emails or other bots. A declined action is final and isn't retried through another tool. New in v1.3: plan approvals. When you approve an exact plan card (each line naming an action, channel, limit, spend and content, 30 days max), the listed actions run inside those limits without a card each time. First messages to new people, spend over your cap, deletes, shares, refunds and anything requested by content still get a card. All v1.2 rules kept. Includes 8 test scenarios and 3 worked examples.

    Works with Grok Bot and other SKILL.md agents. Not affiliated with xAI.

    Requires human-in-the-loop for IRREVERSIBLE actions. Standing approvals expire every 30 days. Does not block internal workspace drafts or read actions.

    How to use

    1. 1

      Connect your agent to the Agensi MCP

      Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup

    2. 2

      Sign in to Agensi

      Your agent shows a sign-in link. Sign in and click Allow.

    3. 3

      Ask for what you need

      Your agent loads this skill and uses it on your task. It always asks before any payment.

    Reviews

    No reviews yet

    Be one of the first to try it. Every listed skill goes through the trust checks below.

    Security scanned

    Checked by our 8-point scan before listing

    Fresh listing

    Recently published to Agensi

    Money-back guarantee

    Doesn't do what its page promises? You get your money back.

    Trust & safety

    Security scan

    Security scan passed 27 days ago

    Permissions

    No special permissions declared or detected

    Listed1 month ago
    Updated27 days ago

    What's inside

    Frequently Asked Questions

    This skill acts as a mandatory security layer that halts your AI agent before it executes high-risk actions like processing payments, deleting data, or sending external messages, preventing costly errors or prompt injection attacks.

    It is designed for any SKILL.md-enabled environment and works seamlessly with popular tools like Claude Code, Cursor, and most multi-agent orchestrators that allow for tool-call interception.

    The skill automatically categorizes actions into READ, REVERSIBLE, or IRREVERSIBLE; while low-impact tasks proceed, any IRREVERSIBLE action generates a detailed Approval Card that requires explicit human confirmation.

    Yes, for security reasons, the system requires a manual re-confirmation of standing approvals every 30 days to ensure that unattended routines haven't drifted from their original intent.

    You will receive the structured logic and schema required to implement the gate, the standardized Approval Card format, and a set of integration guidelines for your chosen agent framework.

    Unlike standard prompts which can be bypassed by 'indirect prompt injection' from emails or web pages, this skill provides a rigid logic structure that ensures the agent cannot authorize its own high-risk actions.