If you discover a security vulnerability in Ledgr, please do not open a public issue.

Instead, report it privately via GitHub Security Advisories or email taniguchi.ryusei@gmail.com with:

1. Description of the vulnerability
2. Steps to reproduce
3. Potential impact
4. Suggested fix (if any)

You will receive a response within 48 hours. We will work with you to understand the issue and coordinate a fix before public disclosure.

The following are in scope:

• Authentication and session management
• Data isolation between households
• Encryption of sensitive credentials (Plaid tokens, AI API keys)
• OAuth flow for MCP server
• Plaid webhook validation
• SQL injection or other injection attacks
• Cross-site scripting (XSS)

The following are out of scope:

• Vulnerabilities in dependencies (report upstream, but let us know)
• Attacks requiring physical access to the host machine
• Self-hosted misconfiguration (e.g., exposing the app without HTTPS)

• All monetary amounts are stored as integers (cents) to prevent floating-point errors
• Plaid access tokens are encrypted at rest at the application layer (AES-256-GCM, authenticated encryption with random per-value IVs). Keys are versioned and rotatable (ENCRYPTION_KEY, ENCRYPTION_KEY_V2, …; see pnpm rotate-keys)
• AI provider API keys are supplied via the AI_API_KEY environment variable and are never persisted to the database
• Other financial data (transactions, balances, investment holdings) is stored in PostgreSQL. Encryption at rest for this data is the responsibility of the deployment — use a managed PostgreSQL instance that encrypts at rest, or a host with full-disk/volume encryption
• Household-based data isolation enforced at the query layer (scopedQuery)
• MCP access is gated behind OAuth with user-granted authorization
• No secrets are stored in the codebase — all credentials come from environment variables
• Transport security (TLS) is provided by the operator's reverse proxy; the application is intended to run behind an HTTPS-terminating proxy and never be exposed over plain HTTP

Ledgr's code, dependencies, and container image are scanned continuously in CI:

• Static analysis (SAST): CodeQL (security-and-quality queries) on every push/PR and weekly.
• Dependency scanning (SCA): Dependabot for npm, GitHub Actions, and Docker base images, with automated update PRs (weekly).
• Secret scanning: GitGuardian on every push/PR.
• Container image scanning: Trivy scans the release image for known OS and library vulnerabilities before it is published. Fixable CRITICAL findings block the release; HIGH/CRITICAL findings are reported to the GitHub Security tab.

Identified, fixable vulnerabilities are remediated on this target timeline: