In development

Rules checked before anything runs

Designed to check each proposed action against written rules before it runs, for defence only.

Sentinel is the policy and defensive-security layer of Ethen Platform. It evaluates whether an action an agent proposes is allowed before the action runs.

Sentinel is part of Ethen Platform and is not sold separately. Its policies are evaluated with Cedar, an open-source policy engine, in development fixtures.

Read the security & trust model Approvals & evidence

Sentinel is defensive only. It is not a penetration-testing or attack tool.

Sample request

Illustration

policy.cedar

Sample written for this page

R1 · permitmatches

permit (
  principal in Role::"research-agent",
  action == Action::"read",
  resource in Folder::"shared-notes"
);

Research agents may read notes in the shared folder.

R2 · forbidno match

forbid (
  principal,
  action == Action::"delete",
  resource in Folder::"records"
);

No one may delete anything in the records folder.

01
Request
A research agent asks to read a note in the shared folder.
02
Who, which action, which resource
principal
Agent::"research-1" (in Role::"research-agent")
action
Action::"read"
resource
Note::"weekly-summary" (in Folder::"shared-notes")
03
Rules evaluated
- R1 permitmatches
- R2 forbiddoes not match
04
Decision
Allow
R1 permits it and no forbid rule matches.

Illustration with a sample policy written for this page, not one of the Platform's fixture policies and not a capture. In Cedar, nothing is allowed unless a rule permits it, and a matching forbid rule always wins.

01Two roles

Someone writes the rules; someone reviews what they do not cover

The design: rules decide where they can, and a person steps in where the policy does not allow an action by itself. The second role is planned.

Role 01Exists in development
Policy author
Writes what agents may and may not do, as rules in the Cedar policy language, checked against a schema.
Today: A policy schema and a set of fixture policies exist. A validation script checks the fixtures against the schema; its results have not been re-run for this page.
Role 02Planned
Reviewer
Looks at an action the policy does not allow on its own, and decides whether it may go ahead.
Today: Planned: connecting policy decisions to the Platform approval service, so a person reviews such an action.

02Decision path

How a policy decision is made

Four steps, from the action an agent proposes to the decision it gets back. The agent acts only on an allow.

Four stepsIllustration

01
Agent proposes an action
An agent working for an Ethen product names the action it wants to take, before taking it.
02
Request: who, which action, which resource
Sentinel builds a request that names who is acting, the action and the resource.
03
Cedar evaluates fixture policies
The Cedar engine, vendored as WebAssembly, evaluates the request against the fixture policies.
04
Allow or deny returned to the caller
The decision, allow or deny, goes back to the caller, which acts only on an allow.

03What exists

Implemented at fixture level, and planned

Sentinel's evidence in the status registry is at fixture level: the policies and the engine exist, but they are not yet applied to real Platform flows.

Implemented

A policy schema and a set of fixture policies written in the Cedar policy language.
The Cedar engine, vendored as WebAssembly with its Apache 2.0 licence kept.
A validation script that checks the fixture policies against the schema. Its results have not been re-run for this page.
Defensive-security agent code carried over from earlier Ethen versions.

Planned

Applying policies to real Platform flows. Policy coverage in real flows is not yet measured.
Connecting policy decisions to the Platform approval service, so a person reviews an action the policy does not allow on its own.

04Limits

Defensive scope, and no guarantee

Defensive scope only. No offensive capability is claimed. Sentinel decides whether an action may run; it does not promise that nothing harmful can happen, and it carries no certification.

Policy evaluation runs in development fixtures only.
Sentinel is defensive only. It is not a penetration-testing or attack tool.
Sentinel is not certified, and it does not make your use of Ethen compliant with any standard.
Not available for general use; nothing here is a release promise or date.

05Questions

Questions

01Does Sentinel carry any certification?
No. Sentinel is not certified, and UpCube holds no compliance certification. The security page explains what we do and do not claim.
02Can I use Sentinel today?
No. Sentinel is in development, and policy evaluation runs only in development fixtures. The availability page shows its current status.
03Is Sentinel an attack or penetration-testing tool?
No. Sentinel is defensive only. It evaluates whether an action an agent proposes is allowed before the action runs, and no offensive capability is claimed.

Sentinel is part of Ethen Platform and is in development. Policies run in development fixtures only.

See availability