threnody-workflowlisted

# Threnody workflow orchestration Runs fan-out work as a tier-aware **Dynamic Workflow** (each `agent()` on its Threnody tier model, not one session model), with multi-queen consensus and learning — then lets you **save a permanent, documented, zero-config `/<slug>`** command teammates run with no setup. This is Claude Code-only. Other hosts use `host_spawn_waves` through `threnody-swarm`; all shared behavior must work there first. ## Fast-start contract Workflow-emitting skills must produce a runnable `workflow_script` quickly: target **under 5 seconds** to handoff and **under 30 seconds** to first worker spawn. The emitted script must start same-wave workers in a batch, for example with `parallel([...])`, before waiting on the wave barrier. Do not block initial workflow emission on optional refinement, consensus, learning aggregation, or permanent-workflow export. Run those after the first worker wave has started or after the workflow returns. For review workflows, use the nested cheap shape: the Workflow script performs medium-tier orchestration, launches file reviewers in parallel, and returns a compact synthesis to the main session. Keep high-tier judgment for explicit deep/security-critical review or validated high/critical findings. ## Prerequisite (claude-code only) This path needs `routing_policy.shells.claude-code.workflow_emit: true` in `~/.local/lib/threnody/config.yaml`. If it is off (or the host is not Claude Code), the response will **not** include `wor