Skip to content

Security: agent-next/cc-manager

Security

SECURITY.md

Security Policy

Scope

cc-manager is a local development tool. It is designed to run on a developer's machine or within a trusted internal network. It has no authentication and should never be exposed to the public internet.

Known Limitations

Area Status Mitigation
Authentication None Local use only; restrict via firewall or reverse proxy
CORS Open (all origins) Restrict at reverse proxy level if needed
Webhook SSRF Partial — blocks private/loopback IPs DNS rebinding not prevented; use trusted endpoints only
Rate limiting Static key ("direct") Does not trust x-forwarded-for; add --trust-proxy if behind proxy

Reporting a Vulnerability

If you discover a security issue, please report it privately:

  1. Do NOT open a public issue.
  2. Email the maintainers or use GitHub Security Advisories.
  3. Include steps to reproduce and potential impact.
  4. We will respond within 7 days.

Security Controls

  • Webhook URL validation: Blocks loopback, RFC 1918 private ranges, and link-local addresses.
  • Rate limiting: 30 requests/minute per client on task submission endpoints.
  • Input validation: Prompt length limits, timeout bounds, priority enum enforcement.
  • Agent isolation: Each agent runs in an isolated git worktree. CLAUDECODE and CLAUDE_CODE_* env vars are cleared to prevent Claude nesting.
  • SQLite WAL mode: Prevents database corruption under concurrent access.

There aren’t any published security advisories