Sealed Write-Only Secret Store

    1

    This store breaks that: sealing a secret needs only the store's public key, so deployment scripts and agents can write all day while being structurally unable to read anything.

    $29

    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 CLIDeepSeekDeepSeekMuseMuse+22 more

    Sealed Write-Only Secret Store

    Example session with this skill installed

    Set up a secret store that our deploy agent can write API keys into but never read, and show me how an operator reads one back.

    • Read your context and instructions
    • Compiled the sealed write-only secret

    Summary

    Initialized a store with a passphrase, created an app token for the deploy agent with a grant on deploy/*, and showed the write/read split.

    What happened

    • node scripts/store.mjs init --db secrets.db printed the store public key; the private key was wrapped under scrypt and the store sealed.
    • The deploy agent, holding only its token, wrote deploy/stripe-key while sealed: HTTP 201. Its attempt to read it back: HTTP 404, identical to a secret that does not exist.
    • An operator unsealed with the passphrase (private key now in memory only), read the value, then sealed again. A wrong passphrase failed the GCM tag and left the store sealed.
    • audit verify walked the hash chain: 6 rows, chain intact, no secret value anywhere in the log.

    Tests

    node --test scripts/store.test.mjs: 7 passed, including the HTTP surface.

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

    About this skill

    Credentials end up scattered across config directories, downloads and chat because the tools that could collect them would also be able to read them. This store breaks that: sealing a secret needs only the store's public key, so deployment scripts and agents can write all day while being structurally unable to read anything. Reading needs the private key, which exists in memory only while an operator has unsealed the store with the passphrase.

    Crypto (crypto.mjs): libsodium's sealed box from Node primitives, X25519 ECDH with an ephemeral sender key, HKDF-SHA-256 with a salt that binds the message key to both the ephemeral and the store key (no replay across stores), AES-256-GCM; the private key wrapped under scrypt (N=2^15) so a wrong passphrase fails the tag and no password hash is stored; hashed tokens; a hash chain for the audit log.

    Store (store.mjs + schema.sql, node:sqlite): initialize once, unseal and seal, write while sealed, metadata listing while sealed, read only when unsealed, rotation by name, per-app tokens shown once and stored hashed, grants with only a trailing *, deactivation, an append-only hash-chained audit that records action, actor and name, never a value, with chain verification that names the first bad row.

    Server (server.mjs): loopback HTTP API; 403 without a token, 404 for an ungranted read identical to a missing secret so the store cannot be enumerated, 423 while sealed, 409 on double initialization, body limits and name validation; admin token from a file or systemd LoadCredential.

    7 tests including the HTTP surface: writes succeed sealed, reads refuse, wrong passphrase stays sealed, cross-store blobs fail, grants and revocation, tampered audit rows detected, no secret value in listings or logs.

    reference/design.md: the security model table, how to operate it, and what it does not protect against. Node 22.5+, no packages, no outbound network.

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

    Requires Node 22.5+. Store is vulnerable if host is compromised while unsealed. Overwrites secret history on rotation. SQLite support may be marked experimental in some Node versions.

    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 19 days ago

    Software
    Node.js 22.5+
    Optional
    systemd

    Node.js needs node:sqlite and node:crypto. systemd is only for LoadCredential-based admin token delivery.

    Permissions

    Terminal / Shell
    Read Files
    Write Files

    Shell to run node scripts and tests; read/write files for the SQLite database file, schema and the admin token file. The server listens on 127.0.0.1 only and makes no outbound network calls.

    Listed19 days ago

    What's inside

    Frequently Asked Questions

    This skill creates a 'unidirectional' vault where agents can deposit API keys or credentials using a public key, but are cryptographically unable to read them back, preventing credential theft via prompt injection.

    It requires Node.js 22.5 or higher to leverage the native `node:sqlite` and `node:crypto` modules, ensuring a zero-dependency footprint for maximum security.

    Yes, you can define specific name patterns and hashed app tokens, allowing you to grant different agents write access to only their specific designated secret paths.

    Each purchase includes the production-ready Node.js skill, an implementation guide for integrating it into your agent's workflow, and a CLI utility for administrative unsealing and auditing.

    Unlike typical stores where 'write' access often implies 'read' access, this uses X25519 sealed boxes to ensure that even a compromised agent cannot be coerced into revealing the vault's contents.

    Secrets are only accessible in plain text when an operator provides the passphrase to unseal the store in memory; once the process ends or the store is sealed, the data remains encrypted on disk.

    New in APIs & Backend