Provenance probably incorrect when creating multiple subjects in one workflow.

[JeroenKnoops]

Problem

When creating multiple subjects in multiple steps in a workflow, all predicates are the same.

Basically the Provenance states: We ran workflow w and it produced subject x with digest y on timestamp z. There is no information on which step and what arguments are used.

Example provenance (from the related issue)

{
  "_type": "https://in-toto.io/Statement/v0.1",
  "predicateType": "https://slsa.dev/provenance/v0.2",
  "subject": [
    {
      "name": "index.docker.io/philipssoftware/blackduck",
      "digest": {
        "sha256": "e4b7db02469ea627818cfd8bcd0cff1c377e64189ef670184e56d0685096374f"
      }
    }
  ],
  "predicate": {
    "builder": {
      "id": "https://github.com/philips-software/docker-blackduck/Attestations/GitHubHostedActions@v1"
    },
    "buildType": "https://github.com/Attestations/GitHubActionsWorkflow@v1",
    "invocation": {
      "configSource": {
        "uri": "git+https://github.com/philips-software/docker-blackduck",
        "digest": {
          "sha1": "64a263c8bb206b256e0507f1998d81e7f647f8fc"
        },
        "entryPoint": "build"
      }
    },
    "metadata": {
      "buildInvocationID": "https://github.com/philips-software/docker-blackduck/actions/runs/1752676715",
      "buildFinishedOn": "2022-01-26T19:39:00Z",
      "completeness": {
        "parameters": true,
        "environment": false,
        "materials": false
      },
      "reproducible": false
    },
    "materials": [
      {
        "uri": "git+https://github.com/philips-software/docker-blackduck",
        "digest": {
          "sha1": "64a263c8bb206b256e0507f1998d81e7f647f8fc"
        }
      }
    ]
  }
}

The invocation node shows which repository with given sha1 and the entryPoint. The entrypoint relates back to the name of the workflow, not the step.
In the metadata we see the buildInvocationID and the buildFinishedOn fields. This is not enough to know what was used to build it. We also need to known what github-action-name is invoked and what arguments were passed.

Possible solution

In the SLSA-provenance v0.2 spec we see an example for "GitHub Actions Workflow".

They are using the entrypoint like we are doing. (they added filename in front of the workflow name). The parameters node is used for workflow_dispatch parameters.
The environment part is to provide context how to reproduce it.. I think we need to add the step / github-action-name and arguments there.

Related issue

See also: philips-software/docker-ci-scripts#97