Skip to content

Security: Quchaosheng/embodied-agent-runtime

Security

SECURITY.md

Security Policy

Supported Code

Security fixes target the current main branch. Older commits and development branches are not maintained as separate release lines.

Reporting a Vulnerability

Do not open a public issue for a vulnerability or include credentials, private network details, robot maps, device identifiers, or deployment logs in public GitHub content.

Report vulnerabilities privately through a GitHub security advisory. Include the affected commit, the software-only reproduction steps, the likely impact, and any suggested mitigation. Remove secrets and site-specific data from attachments.

Credential Handling

  • Keep credentials in environment variables or an external secret store.
  • Never commit .env files, private keys, access tokens, robot credentials, or private service URLs.
  • Use narrowly scoped, revocable credentials for development and CI.
  • Revoke and rotate a credential immediately if it appears in Git history, logs, screenshots, or an issue.

Robot Safety Boundary

This repository demonstrates a deterministic ROS 2 workflow boundary. Its validation, deadlines, cancellation handling, diagnostics, and software vcan0 tests are not a certified functional-safety system and do not replace a hardware emergency stop, safety PLC, drive-level limits, guarded work area, or deployment-specific risk assessment.

Treat every external request, perception observation, network message, and CAN frame as untrusted. Physical camera, CAN, actuator, stopping, X5, BPU/NPU, and GPIO behavior require target-specific validation before use around hardware.

There aren't any published security advisories