EVE Red Team

Run a packaged adversarial regression suite against your own configuration, and prove blocked attacks caused zero unauthorized side effects.

Who this is for

Security and platform teams who want repeatable, adversarial testing of their governance configuration — single-step attacks and multi-step chains — and who need signed results they can keep as regression evidence.

The problem

You can hand-write a few attack cases, but you cannot easily prove that a blocked attack produced no side effect, and you cannot re-run a consistent suite every time the configuration changes. You need a packaged harness that runs safely by default, records what happened, and signs the outcome.

Supported capabilities

• Customer-runnable adversarial suite with signed results. Customers can run a packaged adversarial regression suite (single-step and multi-step chains) against their own configuration; blocked attacks provably cause zero unauthorized side effects, and results are signed (core/redteam/). The reference run covers 16 cases plus 2 chains with 0 unauthorized side effects.

Readiness: the red-team harness is PILOT_READY — customer-pilot ready at supported boundaries.

How it works

1. You configure the harness against your own governance configuration.
2. It runs a representative suite of single-step attacks and multi-step chains.
3. Attacks that should be blocked are blocked, and the harness records that no unauthorized side effect occurred.
4. Results are compiled into a signed report you keep as regression evidence.

The harness is safe by default: local recording targets, no destructive tests, dry-run.

Technical example (CLI)

# Reference invocation against a synthetic/local target — no production credentials
eve redteam run --config ./redteam.config.json --report ./out/redteam-report.json

The eve CLI ships with the EVE service/repository (not the pip client). Recording targets are local; the suite performs no destructive actions.

Signed evidence example

{
  "redteam_report": {
    "passed": true,
    "cases": 16,
    "chains": 2,
    "unauthorized_side_effects": 0,
    "standards_mappings": ["descriptive-only"],
    "signature": "ed25519-…",
    "canon": "jcs-1"
  }
}

Standards mappings are descriptive, not certification.

Verifier example

from core.eve_sdk import verify_evidence

report = verify_evidence("redteam", redteam_report)
assert report["valid"] is True

Deployment options

The red-team harness runs embedded inside the EVE service or as a sidecar, against your own configuration. Signed reports verify offline in Python (and, for verify-only, TypeScript).

Readiness

The red-team harness is PILOT_READY — customer-pilot ready at supported boundaries. Signed reports use jcs-1 and offline verification, which are SUPPORTED. The EVE SDK core overall is a RELEASE CANDIDATE WITH EXCLUSIONS.

Limitations

• The harness is safe by default: local recording targets, no destructive tests, dry-run.
• The suite is representative, not exhaustive — customers add their own fixtures.
• Standards mappings are descriptive and are not certification.
• The eve CLI ships with the service/repository, not the pip client.
• The release-candidate clients are distributed privately for pilots — install the pilot artifact, not a public pip/npm install (see readiness for public-registry status).

Next step

Request a pilot to run the packaged suite against your configuration and keep the signed report as adversarial regression evidence.