More screenshots

    Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIDeepSeekDeepSeekMuseMuse+22 more

    Make Your Agents Ask Before They Act

    1

    Two checks before an agent acts: proof of which agent is asking, and a person's signed yes to the exact action — amount, window, everything.

    $15

    Secure checkout via Stripe

    Money-back guarantee

    Converts to your local currency at checkout

    Security scanned

    Make Your Agents Ask Before They Act

    Example session with this skill installed

    Generate a proof request for a $50.00 cloud hosting payment using the x401 protocol and prepare the mandate for my approval wallet.

    • Read your context and instructions
    • Compiled the make your agents

    I have generated the x401 proof request and prepared the scoped mandate for your review. The challenge has been validated, and the mandate is now pending in your local approval queue.

    Scoped Mandate and x401 Headers

    Mandate Details (Pending Approval):

    FieldValue
    Verifiercloud-provider-01
    Actioninfrastructure.payment
    Amount Ceiling"50.00"
    CurrencyUSD
    Nonce8f2d3a9b1c5e4f7g8h9i0j1k2l3m4n5o
    Expires2023-11-01T15:45:00Z

    Generated PROOF-REQUEST Header:

    PROOF-REQUEST: alg=EdDSA, kid=v-key-882, nonce=8f2d3a9b1c5e4f7g8h9i0j1k2l3m4n5o, exp=1698853500, iss=cloud-provider-01
    

    Canonical JSON for Signing:

    {
      "agent": "deploy-bot-v1",
      "amt": "50.00",
      "cat": "infrastructure.payment",
      "exp": 1698853500,
      "nbf": 1698849900,
      "nonce": "8f2d3a9b1c5e4f7g8h9i0j1k2l3m4n5o",
      "ver": "cloud-provider-01"
    }
    

    Next steps

    1. Open your local approval wallet at http://localhost:8080/approve/772-910.
    2. Review the canonical JSON terms and verify the amount ceiling.
    3. Sign the mandate to generate the PROOF-RESPONSE JWS token.

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

    About this skill

    Two questions get blurred together whenever an agent acts: who is behind this agent, and did a person approve what it's about to do? This skill separates them and answers both. A signed challenge-and-proof header protocol (called x401) answers the first. An approval wallet that runs on the person's own machine answers the second — the person reviews the exact terms on their own device and signs a narrow permission, with no commercial identity network and no central registry in the loop.

    Inside the package: the skill file covering the protocol, replay protection, the wallet design, and the trust model; an example set of mandate terms — the exact claims a person sees before signing; and an approval-flow checklist. Approval is never open-ended: the signed mandate names the agent, the other party, one action category, an amount ceiling (carried as text, never a decimal number, so rounding can't change it), and a not-before/not-after window of about an hour. Change a single character after signing and the signature fails.

    A run goes like this: the agent requests a resource and gets back a signed challenge carrying a fresh single-use code. It submits the action's terms to the wallet, which fixes those exact terms at submission time. The person sees who is asking, what's requested, the ceiling, the window, and the exact data their key will sign — then approves or denies. Pending approvals expire in about ten minutes; an expired one can never be approved. On approval, the wallet signs and the agent forwards the signed mandate as its proof. Trust is a local list the operator owns, empty by default — nothing is trusted until you add it. Scope, honestly: this package is the complete design and rules, written to be built from. It is a specification and method, not a finished wallet program.

    Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for Security.

    Requires Ed25519 support. No support for auto-approval in production. Mandates must use string-based amounts. Localhost binding required for wallet security.

    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 buying a skill.

    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 6 days ago

    An Ed25519 keypair and a self-hosted approval wallet.

    Permissions

    Terminal / Shell
    Read Files
    Write Files
    Network Access
    Environment Variables
    Listed6 days ago

    What's inside

    Frequently Asked Questions

    This skill bridges the trust gap by implementing x401 headers for cryptographically signed identity and a human-in-the-loop wallet, preventing agents from performing unauthorized 'blank cheque' actions.

    This skill is compatible with any AI agent framework that supports custom HTTP headers and external service calls, as it operates at the protocol level using Ed25519 JWS standards.

    The purchase includes the core x401 identity protocol implementation, the self-hosted human approval wallet interface, and the cryptographic logic for mandate generation and nonce management.

    Yes, for security and precision, the skill strictly uses string-encoded values for mandates to avoid the floating-point rounding errors common in LLM-generated math.

    The skill uses device-local Ed25519 private keys stored with owner-only file system permissions, ensuring that your sensitive signing keys never leave your controlled hardware.

    By default, the skill enforces a human-in-the-loop requirement for all sensitive actions to ensure maximum security; it does not support fully autonomous or 'auto-approval' paths.

    New in Security