Do not open a public GitHub issue for security vulnerabilities.

Email: security@medguard.dev

Please include:

• Description of the vulnerability
• Steps to reproduce
• Potential impact
• Suggested fix (if any)

We will respond within 48 hours and aim to release a fix within 7 days for critical issues.

• Agent keypair is stored in .env and never committed to version control
• Spending policy is enforced locally before every payment — the LLM cannot bypass it
• Approval threshold blocks payments above a caregiver-set limit
• All Stellar transactions are publicly verifiable on stellar.expert
• No user PII is stored — only spending logs and wallet addresses

• This is a testnet application. Do not use mainnet funds without a full security audit.
• The LLM system prompt is the primary guardrail — a sufficiently adversarial prompt could attempt to manipulate the agent. The policy engine provides a hard limit.
• JSON file storage is not suitable for production multi-user deployments.