Skip to content

Security: Jaival-Suthar/agentguard

Security

SECURITY.md

Security Development Rules

AgentGuard is an assurance layer that observes agent execution. The security boundary matters because the runtime can invoke tools, execute code, and access external services.

Trust Boundary

TrueForge execution
    ↓
AgentGuard verification
    ↓
authoritative assurance artifact
    ↓
Assurance Console visualization

The UI is not trusted to decide PASS or FAIL. The authoritative decision comes from AgentGuard.

Local Development Rules

  • Use isolated local services for the experimental stack.
  • Use synthetic incident data only.
  • Do not mount the host home directory into containers.
  • Do not expose SSH keys, browser profiles, cloud credentials, Git credentials, or .env files to agent-facing processes.
  • Do not run arbitrary agent-generated commands directly on the Windows host.
  • Keep TrueForge bound to the intended localhost / host-published port.
  • Use explicit, minimal permissions for MCP tools and sandbox resources.

Docker Boundaries

The repository includes an optional compose-based AgentGuard API and Assurance Console path.

  • agentguard-api listens on container port 8780 and is published to the host on 8780
  • assurance-ui serves the built console on host port 5174
  • agentguard-runner is an optional profile that connects to an already-running TrueForge host service through the host-gateway mapping

The compose path is implemented, but full end-to-end Docker validation is still pending in this documentation pass.

Secrets And Environment Variables

Environment files are local-only configuration. Never commit real secrets.

The current repository uses:

  • TRUEFORGE_BASE_URL
  • TRUEFORGE_MODEL_NAME
  • TRUEFORGE_AGENT_NAME
  • TRUEFORGE_INCIDENT_ID
  • TRUEFORGE_MCP_SERVER_NAME
  • DAYTONA_API_KEY for sandbox configuration only

The .env.example file is a documentation aid, not a secret store.

Data Boundaries

  • data/runs/ contains raw run metadata and JSONL evidence
  • data/assurance/ contains generated assurance artifacts
  • the live console can import artifacts, but the UI is not the source of truth

Security Goal

AgentGuard should verify execution authority and policy invariants from observed runtime evidence rather than from model narration or UI state.

Verification Expectation

Any security-sensitive change should be accompanied by:

  • automated tests
  • deterministic verification
  • runtime inspection of the generated evidence

There aren't any published security advisories