Security Policy

Supported Versions

Version	Supported
0.1.x	✅

If you discover a security vulnerability in Deerflow, please report it responsibly:

Do not open a public GitHub issue.
Report the vulnerability through GitHub Security Advisories or contact the maintainer directly via GitHub.
Include the following in your report:
- A description of the vulnerability
- Steps to reproduce
- Potential impact
- Any suggested mitigations (optional)
Allow up to 72 hours for acknowledgment before any public disclosure.

We take all security reports seriously and will respond promptly.

When deploying Deerflow, follow these security practices:

Never commit secrets to version control. Use environment variables for all sensitive values.
Copy .env.example to .env.local and fill in your own values. Never commit .env.local.
Generate strong, unique secrets for JWT_SECRET and DEERFLOW_LOCK_SECRET using a cryptographically secure random generator.
In production, set CORS_ORIGIN to your specific domain rather than using wildcard (*).
Run npm audit --audit-level=moderate regularly and before every deployment.
Keep dependencies updated. Enable Dependabot for automated security updates.

Deerflow includes several built-in security enforcement mechanisms:

Feature	Module	Description
Deep Security Scan	`deerflow/enforcement/security-deep.ts`	Scans for OWASP Top 10 and CWE patterns (eval, innerHTML, injection)
npm Audit Gate	`scripts/quality-gate.ts`	Blocks deployment on moderate+ vulnerabilities
Anti-Pattern Detection	`deerflow/enforcement/anti-pattern.ts`	Flags eval(), innerHTML, hardcoded secrets, and other dangerous patterns
Fabrication Detection	`deerflow/enforcement/fabrication-detector.ts`	Catches code referencing non-existent security packages
Tamper-Proof Lock	`deerflow/enforcement/tamper-proof-lock.ts`	SHA-256 hash chain prevents bypassing enforcement gates

The tamper-proof lock secret is derived from environment variables or a machine fingerprint. On shared CI machines, set the DEERFLOW_LOCK_SECRET environment variable for consistent behavior.
The fabrication detector uses pattern matching rather than full AST analysis. It may produce false positives or miss novel fabrication patterns.
The security deep scan covers common vulnerability patterns but does not replace a dedicated SAST tool such as Semgrep or CodeQL.

Security updates are published as patch releases. We recommend enabling Dependabot or Renovate for automatic security dependency updates.