Improve a Queue

What this skill is

This is the guided improvement loop for a Queue and the workflow around it: the conversational shell that takes the user from "this Queue could work better" to a change that is actually applied. A Queue is rarely improved alone — it sits between a producer that pushes cases in and a consumer triggered to work them, and most of what makes it slow, inconsistent, or backed-up lives in that structure and in the AOPs at the seam between the two Agents.

You own the loop and the judgement calls — which Queue, what to base the improvement on, which Runs and cases to look at, what to change, whether to apply it. You do not re-derive the analysis or the AOP rewrites yourself: workflow-debugger already knows how to audit a producer→consumer workflow from its topology, backlog, and the AOPs at the seam, and aop-writer already knows how to rewrite an AOP. You orchestrate them.

The value you add over running workflow-debugger directly is the loop around it: pinning the Queue, surveying its full shape, getting the user to choose the basis, proposing and confirming a range, and closing the loop by applying what they accept.

Where this sits among the skills

• workflow-debugger — the read-only audit of a Queue workflow: topology and problems, the backlog, both producer and consumer AOPs, and a sample of Runs. You call it for the analysis in step 5; you don't reproduce its taxonomy here.
• run-debugger — transcript-level diagnosis of one Run. Reach for it when a recurring failure (a case that won't close, a wrong branch) needs depth on a representative Run.
• aop-writer — rewrites an AOP from the in-effect AOP plus a change request. Every AOP change you apply — producer or consumer — goes through it; you never rewrite an AOP inline.

If the user wants one standalone Agent improved rather than a Queue-connected workflow, that's the improve-agent skill.

The loop