Skip to content

Security: QuantumClaw/QClaw

Security

SECURITY.md

Security Policy

Reporting a Vulnerability

Do not open a public issue for security vulnerabilities.

Email security@allin1.app with:

  • Description of the vulnerability
  • Steps to reproduce
  • Potential impact
  • Suggested fix (if you have one)

We will acknowledge receipt within 48 hours and aim to provide a fix within 7 days for critical issues.

Supported Versions

Version Supported
1.x Yes

Security Architecture

QuantumClaw takes security seriously at every layer:

  • Secrets: AES-256-GCM encryption at rest, machine-specific derived keys
  • Trust Kernel: Immutable VALUES.md rules the agent cannot override
  • Audit Trail: Every agent action logged with timestamps and cost tracking
  • AGEX Protocol: Autonomous credential lifecycle with automatic rotation, scoped delegation, and emergency revocation in under 60 seconds
  • Exec Approvals: Destructive operations require human sign-off
  • Channel Allowlists: Only authorised users can interact with your agent

Known Limitations

  • The self-signed development AID (ia_signature: 'self-signed-dev-not-for-production') is for local development only. Production deployments should use a certified AGEX Identity Authority.
  • The local dashboard binds to localhost by default. Do not expose it to the internet without authentication.
  • The completion cache stores prompt hashes, not plaintext prompts, but cache entries could theoretically reveal usage patterns.

Acknowledgements

We appreciate responsible security researchers. Contributors who report valid vulnerabilities will be credited (with permission) in our changelog.

There aren’t any published security advisories