Skip to content

Security: claude-world/agentOS

Security

SECURITY.md

Security Policy

Supported Versions

Version Supported
1.7.x Yes
< 1.7 No

Reporting a Vulnerability

Please do NOT open public issues for security vulnerabilities.

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

  1. Email: Send details to security@lucas-futures.com
  2. Subject: [AgentOS Security] Brief description
  3. Include:
    • Description of the vulnerability
    • Steps to reproduce
    • Potential impact
    • Suggested fix (if any)

Response Timeline

Action Timeline
Acknowledgment Within 48 hours
Initial assessment Within 5 business days
Fix development Depends on severity
Public disclosure After fix is released

Scope

The following are in scope for security reports:

  • API key exposure — Any path where API keys could leak (logs, crash reports, network traffic)
  • Channel authentication bypass — Circumventing Telegram/Discord/LINE allow-lists
  • Tool execution without consent — Bypassing PrivacyHook for sensitive operations
  • MCP server spoofing — MITM or injection via MCP connections
  • Data exfiltration — Unauthorized access to on-device data (photos, contacts, files)
  • Relay worker vulnerabilities — Issues in the Cloudflare Worker relay infrastructure
  • Webhook verification bypass — Bypassing Ed25519/HMAC-SHA256 signature verification

Out of Scope

  • Vulnerabilities in third-party services (Anthropic API, Telegram API, etc.)
  • Issues requiring physical access to an unlocked device
  • Social engineering attacks
  • Denial of service via rate limiting exhaustion (rate limiter is already in place)

Security Architecture

AgentOS follows these security principles:

  1. Keychain-only storage — API keys and tokens are stored exclusively in iOS Keychain
  2. Fail-closed channel auth — Channels reject messages from non-allowlisted sources by default
  3. HTTPS-only MCP — No plaintext MCP connections allowed
  4. PrivacyHook gating — Sensitive tools (Drive uploads, Gmail sends, GitHub writes, contacts, BLE, HomeKit) require explicit user permission
  5. ConstitutionHook — Constitutional AI guardrails prevent harmful tool use patterns
  6. Webhook verification — All incoming webhooks are cryptographically verified
  7. Token-bucket rate limiting — Prevents API abuse

Recognition

We appreciate security researchers who help keep AgentOS safe. Contributors who report valid vulnerabilities will be acknowledged in release notes (with permission).

There aren’t any published security advisories