Skip to content

Security: bayshanhai-dev/agentweave

Security

SECURITY.md

Security Policy

Supported versions

Until the first stable release, security fixes are applied to the default branch. Published release support will be documented here when versioned releases begin.

Reporting a vulnerability

Please do not disclose a suspected vulnerability in a public issue. Open a private security advisory through GitHub, or contact the project maintainers privately with the subject AgentWeave security report.

Include:

  • affected commit, release, or component;
  • reproduction steps or a minimal proof of concept;
  • impact and any required permissions;
  • suggested mitigation, if known.

We will acknowledge reports as soon as practical, investigate, coordinate a fix and disclosure timeline, and credit reporters who want attribution.

Deployment guidance

AgentWeave is currently a single-host MVP. Before exposing it beyond a trusted local network:

  • protect the Control API and dashboard with an authenticated reverse proxy;
  • keep PostgreSQL, NATS, and observability ports private;
  • use strong, unique provider and bridge credentials;
  • rotate a bridge token immediately if it appears in a terminal transcript, screenshot, issue, or shared log;
  • set a narrow workspace allowlist and never mount a broad home directory;
  • do not place credentials in .env committed to Git;
  • review logs before sharing them, because prompts and tool output may contain sensitive project information;
  • keep Docker, Node.js, PostgreSQL, NATS, and provider clients patched.

Provider credentials must remain in runtime configuration and must not be persisted in event payloads, task evidence, or dashboard read models.

There aren't any published security advisories