Set up Attack Harness
Install the released container image into the local Docker daemon used by Operator. Operator then supplies campaign inputs and manages startup, execution, and shutdown.
Obtain the released image
Start at official releases, the
public repository, and its
container package.
Choose your documented release and linux/amd64 or linux/arm64 matching Docker.
From your host shell, set the version and platform explicitly:
HARNESS_VERSION=0.1.5
HARNESS_PLATFORM=linux/arm64
docker pull --platform "$HARNESS_PLATFORM" "ghcr.io/intrusiveai/attack_harness:v${HARNESS_VERSION}"
docker image inspect "ghcr.io/intrusiveai/attack_harness:v${HARNESS_VERSION}" --format '{{json .RepoDigests}}'
For an immutable selection, use the released image/index digest from the catalog, pull it explicitly, and record the resolved platform image identity:
HARNESS_IMAGE=ghcr.io/intrusiveai/attack_harness@sha256:RELEASED_DIGEST
docker pull --platform "$HARNESS_PLATFORM" "$HARNESS_IMAGE"
Replace RELEASED_DIGEST with the published value; it is not an example you can run
unchanged. Operator does not pull the image for you.
Select it in Operator
Set Operator's engine.image to the local reference. Its installed contract version
and exact AsyncAPI document digest must match the image's embedded compatibility
record. Product version labels and image archive checksums are different identities.
Operator verifies the actual image during admission; never edit compatibility files
to bypass a mismatch.
Follow Operator setup for model access, host configuration, and target selection, then campaign usage. The links identify the 0.1.5 example; use your installed version's documentation for other release selections.
The selected release's integration guide and contract lock provide deeper compatibility details. Do not manually run the image as a replacement for Operator's campaign workflow.