bagman

Secure AI agent access to private keys and API secrets.

23|5|Updated Feb 23, 2026
One-click install
npx skills add https://github.com/EverClaw/EverClaw --skill bagman-everclaw
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: bagman
Source: https://github.com/EverClaw/EverClaw/tree/main/skills/bagman
Command: npx skills add https://github.com/EverClaw/EverClaw --skill bagman-everclaw

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve?

Bagman protects AI agents that need access to wallets, API keys, credentials, and other sensitive secrets. It prevents accidental disclosure, blocks prompt-injection attempts, and replaces master-key exposure with safer delegated access patterns.

Core Features & Use Cases

  • Secure secret retrieval: Pull credentials at runtime from 1Password, macOS Keychain, encrypted files, or environment-backed fallbacks.
  • Leak prevention: Sanitize every outbound response, cron summary, and debug message so keys, tokens, seed phrases, and sensitive operational metrics do not escape.
  • Agent safety controls: Validate user input, block jailbreaks, require confirmations for risky actions, and enforce allowlisted wallet operations.
  • Session-based access: Use bounded ERC-4337-style session keys with expiration, spending limits, and contract restrictions instead of master keys.
  • Use case: An autonomous trading agent can fetch a short-lived session key, execute only approved swaps, and revoke access cleanly if behavior becomes suspicious.

Quick Start

Ask the AI to secure your agent workflow by storing secrets in a supported backend, enabling input validation and output sanitization, and creating bounded session keys for any wallet operations.

Frequently Asked Questions about bagman

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I secure AI agent access to wallet credentials without exposing raw secrets?

Secure AI agent access by retrieving credentials at runtime from backends like 1Password or macOS Keychain, enforcing strict allowlists, and replacing master-key exposure with bounded, revocable session keys.

Can I prevent prompt-injection attacks from leaking API keys during autonomous workflows?

Prevent prompt-injection leaks by validating user input, blocking jailbreaks, and sanitizing every outbound response, cron summary, and debug message to ensure tokens and seed phrases do not escape.

How do I set up session keys for autonomous trading agents to limit financial risk?

Set up ERC-4337-style session keys with expiration dates, spending limits, and contract restrictions, allowing autonomous trading agents to execute approved swaps and revoke access cleanly if behavior becomes suspicious.

Does this approach work with 1Password and macOS Keychain for runtime secret retrieval?

Runtime secret retrieval works with 1Password, macOS Keychain, encrypted files, or environment-backed fallbacks, pulling credentials dynamically at runtime instead of hardcoding them into agent workflows.

What is the best way to sanitize outbound AI responses to prevent sensitive operational metrics from leaking?

Sanitize outbound AI responses by applying leak prevention filters to every debug message and cron summary, ensuring keys, tokens, seed phrases, and sensitive operational metrics are scrubbed before they escape.

When should I not use master keys for AI agent payment automation?

Avoid using master keys for payment automation when you need auditable operations; instead, require confirmations for risky actions and enforce allowlisted wallet operations using bounded session-key permissions.