Missing Threat Detection Custom Pre-Steps

[chrisfregly]

Background

Currently gh-aw supports wrapping agentic execution in custom steps by prepending via a steps field and appending via a post-steps field. These steps are then added to the agent job when a workflow is compiled.

Request

With the addition of threat detection into gh-aw we now have a new detection job that also provides agentic execution, just in this case for security scanning. However unlike the main agent job, threat detection supports a steps field that adds custom workflow steps only after agentic execution. It would be ideal if the detection job supported both pre and post steps similar to the agent job. Ideally we would follow a similar schema where there are two fields, steps and post-steps that surround the agentic execution of this detection job.

Use Case

We are building out agentic workflows that connect to models behind a private AI gateway. As a pre-step, we do some extra work to securely connect to this gateway. Currently we can do this extra connection work in agent jobs thanks to the steps field and would also like to follow a similar pattern when using this detection job.