refinery-mergelisted

# Refinery Merge Sequential merge queue processor for completed polecat branches. The refinery is the DMA controller of the Gastown chipset -- it moves data (code changes) from producer (polecat branches) to consumer (main branch) without CPU (mayor) intervention during normal operations. The pipeline is deterministic: checkout, rebase, test, merge, push. Merge conflicts block the queue and escalate; they are never auto-resolved. ## Activation Triggers This skill activates when: - The merge queue has pending merge requests from completed polecats - A polecat calls done and its branch needs to be merged into main - The refinery is assigned to process a rig's merge queue - Merge queue depth exceeds a threshold and processing is needed ## Core Capabilities ### Sequential MR Processing The refinery processes merge requests one at a time in strict FIFO order. No parallel merges. No priority reordering. The queue is a pipeline, not a pool. **Queue processing loop:** ```typescript const state = new StateManager({ stateDir: '.chipset/state/' }); async function processQueue(): Promise<void> { // Get pending MRs in FIFO order const pending = await listMergeRequests('pending'); if (pending.length === 0) { // Queue empty -- nothing to process return; } // Process the oldest MR first (strict FIFO) const mr = pending[0]; await processMergeRequest(mr); } ``` ### Rebase-Test-Merge Pipeline Each merge request flows through a five-stage pipeline. Every stage