Skip to content

Security: Nigmat-future/omp-route

Security

SECURITY.md

Security Policy

omp-route is a local planning tool. It reads OMP's model catalog and a redacted usage report, writes a configuration overlay, and launches the OMP binary. It never contacts model-provider APIs and never proxies model traffic.

Threat model

What the tool is designed to guarantee:

Property Mechanism
No credential access Usage is requested with OMP's --redact flag; error output is additionally scrubbed for API-key and bearer-token patterns before it is printed
No network egress The planner only spawns the locally installed OMP CLI; there are no HTTP clients and no runtime dependencies
No shell injection OMP is spawned without a shell, and passthrough arguments are forwarded as an array
Reversible writes The overlay is written to its own file (~/.omp-route/routes.generated.yml, mode 0600 where supported); the user's OMP configuration is never edited
Fail closed Invalid model-discovery JSON aborts before OMP is launched
Bounded automation Unscoped plans above 200 routes are refused unless --all is passed

Out of scope: the behavior of OMP itself, the model providers it talks to, and the correctness of a route once OMP has taken over execution.

Supported versions

omp-route is a proof of concept. Only the main branch receives fixes.

Reporting a vulnerability

Please report suspected vulnerabilities privately through GitHub's private vulnerability reporting rather than in a public issue.

Useful details: the command you ran, the OMP version, the generated overlay (with any provider identifiers you consider sensitive removed), and what you expected instead. A response can be expected within 14 days.

If the issue is a leak of secrets into output, please include the redaction path that failed — that code lives in src/omp.js.

There aren't any published security advisories