Skip to content

Security: protagolabs/NarraNexus

Security

SECURITY.md

Security Policy

Supported versions

We support the latest minor release of NarraNexus. Older minor versions do not receive backported fixes — please upgrade.

Version Supported
Latest v1.x
Earlier v1.x

Reporting a vulnerability

Please do not open a public issue for vulnerabilities.

Use GitHub's private vulnerability reporting:

Report a vulnerability →

Or email the maintainers directly — see MAINTAINERS.md for the current contact address.

What to include

A useful report contains:

  • The component affected (e.g. backend/routes/auth.py, xyz_agent_context.agent_runtime, the Tauri sidecar, …).
  • A minimal reproduction — request payload, env config, or steps.
  • Impact: what an attacker can read, write, or trigger.
  • Suggested fix if you have one (not required).

Response targets

  • Acknowledgement within 72 hours of report.
  • Initial assessment within 7 days.
  • Remediation plan within 14 days for high-severity issues (auth bypass, RCE, secret leakage). Lower-severity issues are triaged on the normal release cadence.

We will keep you in the loop and credit you in the release notes if you wish.

In-scope areas

  • Authentication & session handling (backend/routes/auth.py, backend/auth.py)
  • Identity resolution in API routes (X-User-Id / JWT)
  • Bundle export — what credentials and chat content can leave (src/xyz_agent_context/bundle/)
  • LLM prompt-injection paths that lead to data exfiltration or unauthorized tool calls
  • Local-mode WebSocket auth (/ws/agent/run)
  • Tauri sidecar privilege escalation or path escapes

Out of scope

  • Issues that require a malicious local OS user with file-system access — local mode's security boundary is the OS user, by design
  • Volumetric denial of service against a self-hosted instance
  • Findings in third-party dependencies without a NarraNexus-specific amplification path — please report those upstream
  • Findings against the cloud deployment at agent.narra.nexus unrelated to the open-source code (operational issues belong to the hosting team, not this repo)

Disclosure

We follow coordinated disclosure. After a fix lands and users have had a reasonable window to upgrade, we publish a brief advisory summarizing the issue, the fix, and any required user action.

There aren't any published security advisories