Skip to content

Security: N-JrP/agentops-local

Security

docs/security.md

Safety and permissions

  • Current tools are registered as read-only.
  • Unknown tools are rejected before execution.
  • Tool input length and null-byte checks are enforced centrally.
  • A narrow prompt/tool-abuse guard blocks direct attempts to bypass system/tool policy.
  • The graph contains a human-approval branch for future tools marked write.
  • No write-capable external tool is enabled in the portfolio build, so no destructive action is possible through the current tool registry.

There aren't any published security advisories