Case · DiggingBeagle record

OmniRoute patched skill-write routes that did not consistently demand management authority

The November 14 changes added explicit management authorization to endpoints generating agent skills and installing marketplace skills. Before this change, the deployment option disabling dashboard login could leave filesystem-changing routes without equivalent admin enforcement.

Public disclosure: Nov 14, 2026

Scope

An OmniRoute service exposes management endpoints while dashboard login is disabled or otherwise inconsistently enforced.

UnratedAI role: AGAINST AIAssessment method

At a glance

Mechanism and trust boundary

  1. 01

    Starting condition

    An OmniRoute service exposes management endpoints while dashboard login is disabled or otherwise inconsistently enforced.

    Boundary: precondition and attacker capability

  2. 02

    Trust boundary crossed or evaluated

    The handlers write or replace SKILL.md and related shared agent capabilities. That crosses from an HTTP caller into code and instructions subsequently offered to agent workers; this is more consequential than a public read-only catalog endpoint.

    Boundary: mechanism

  3. 03

    Observed consequence

    A caller without management authorization could reach skill generation or installation actions that change future agent capabilities.

    Boundary: observed effect and limitations

Claims & evidence

CLM-NOV15-OMNISKILL-MECHThe November 14 changes added explicit management authorization to endpoints generating agent skills and installing marketplace skills. Before this change, the deployment option disabling dashboard login could leave filesystem-changing routes without equivalent admin enforcement. The handlers write or replace SKILL.md and related shared agent capabilities. That crosses from an HTTP caller into code and instructions subsequently offered to agent workers; this is more consequential than a public read-only catalog endpoint.supported

Basis: reported finding

Link to claim
CLM-NOV15-OMNISKILL-SCOPEThe commit establishes the intended fix and added tests, not a demonstrated anonymous production compromise or a guaranteed packaged release. Exact exposure still depends on deployment configuration and routing. The evidence is bounded by the cited source and test setting. Broader incidence, other environments, and actual production consequences are not established merely by these observations.supported

Basis: reported finding

Link to claim

Implications

A caller without management authorization could reach skill generation or installation actions that change future agent capabilities.

Controls and mitigations

  • Require management-scoped API credentials on every write route regardless of dashboard login mode.
  • Review newly installed skills and runtime capabilities against an immutable, administrator-approved inventory, and test anonymous/inference-only callers after upgrading.

Unknowns and contradictions

  • The evidence is bounded by the cited source and test setting. Broader incidence, other environments, and actual production consequences are not established merely by these observations.

Sources and citation

Material revision history

  1. Nov 15, 2026 · Canonical change recorded · new in release · revision 250