Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIMuseMuseDotsDots+21 more

    agentsec πŸ”’

    1

    Securely store, inject, and audit secrets for AI agents using an MCP-compatible encrypted vault.

    Free

    Security scanned
    agentsec πŸ”’

    agentsec πŸ”’

    Example session with this skill installed

    List all secrets in the current namespace and then run a scan to check for leaks in the workspace.

    • Read your context and instructions
    • Compiled the agentsec

    I have retrieved the secret metadata for the current namespace ws:agentsec-dev and performed a security scan of the workspace. No active secrets were detected in the local files.

    Secret Metadata and Scan Results

    Key NameDescriptionTTL RemainingStatus
    STRIPE_API_KEYProduction payment gateway24,521sSecure
    AWS_SECRET_ACCESS_KEYCI/CD deployment userPermanentSecure
    DATABASE_URLLocal postgres connection3,600sSecure
    GITHUB_TOKENRepo management86,400sSecure

    Scan Summary

    • Files Scanned: 42
    • Matches Found: 0
    • Namespace: ws:agentsec-dev
    • Daemon Status: DAEMONLESS (Direct Local Vault)

    Next steps

    • Run agentsec doctor to verify vault integrity and namespace permissions.
    • Use agentsec run --only STRIPE_API_KEY -- npm run deploy to use credentials safely.
    • Configure agentsec renew for the DATABASE_URL if the task requires more than one hour.

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

    About this skill

    Managing credentials in plaintext .env files or CLI arguments risks leaking secrets to logs, commits, and AI transcripts. Agentsec provides a secure vault with an MCP interface to store, inject, and audit secrets without exposing them to the agent's context.

    What it does

    • Secret Injection: Injects credentials directly into process environments so they never appear in command strings or logs.
    • Automated Redaction: Redacts injected values from stdout and stderr automatically to prevent accidental display in chat transcripts.
    • Leak Scanning: Scans workspaces for plaintext matches of stored secrets before commits or deployments.
    • Namespace Management: Segregates secrets by repository or directory with granular access controls and individual file locks.
    • Audit Logging: Tracks access metadata including client name, operation type, and outcome for security monitoring.

    How it works

    1. Initialize Vault: Run agentsec setup to configure the vault and install necessary MCP adapters for your agent environment.
    2. Store Secrets: Use set_secret or agentsec set to move keys from environment files into the encrypted vault.
    3. Execute Tasks: Run commands via run_command_with_secrets to provide only the specific keys a process requires.
    4. Audit and Scan: Periodically run agentsec audit to review access or agentsec scan to check for leaks in the codebase.

    Frameworks & tools

    Built for agents using MCP stdio or HTTP interfaces. Works with Codex, Grok, Gemini CLI, and Antigravity. Supports local execution via CLI with fallback to daemonless mode using encrypted namespace files.

    Why this beats prompting it yourself

    Manual prompting cannot prevent an agent from printing a secret to the console or including it in a shell command. Agentsec handles the encryption, injection, and redaction at the system level, ensuring sensitive data remains invisible to the LLM.

    Use cases

    • Injecting API keys into deployment scripts without writing them to disk.
    • Rotating master tokens and renewing short-lived leases for temporary access.
    • Scanning local repositories for leaked credentials before pushing to remote.
    • Restricting agent tools to specific secret namespaces based on the current project.

    Known limitations

    Scanning only matches stored values of six characters or more and is not a complete credential detector. Redaction does not cover encoded, transformed, or partial output.

    Scanning requires at least six characters to match. Redaction does not cover encoded or transformed values. Requires local CLI installation or daemon access.

    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

    Free forever

    No account required to browse

    Trust & safety

    Security scan

    Security scan passed 16 days ago

    • Free with an Agensi account

    Requires

    macOS/Linux/Windows/BSD
    Git
    macOS

    Permissions

    Terminal / Shell
    Write Files
    Network Access
    Read Files
    Environment Variables

    The tool can be modified freely so if you think there is a better way then ask your agent to improve it

    Listed16 days ago

    What's inside

    Frequently Asked Questions

    Agentsec acts as a secure intermediary that injects credentials directly into the environment during runtime, preventing secrets from ever appearing in command history, logs, or the LLM's own chat context.

    It is designed for any agent using the Model Context Protocol (MCP) via stdio or HTTP interfaces, including platforms like Codex, Grok, and Gemini CLI.

    Installation is straightforward: run the 'agentsec setup' command to configure the encrypted vault and install the necessary MCP adapters for your specific environment.

    Your purchase includes the core MCP-compatible encrypted vault, the CLI management tool, automated redaction scripts, and the leak scanning utility.

    The scanning feature performs a plaintext match against your stored secrets to identify accidental exposures in your workspace before you commit or deploy code.

    Yes, you can use namespace management to segregate secrets by project or directory, ensuring an agent only has access to the specific keys required for its current task.

    New in LLM Ops