diff --git a/.story_kit/work/1_upcoming/230_story_prevent_duplicate_stage_agents_on_same_story.md b/.story_kit/work/1_upcoming/230_story_prevent_duplicate_stage_agents_on_same_story.md
new file mode 100644
index 0000000..f304936
--- /dev/null
+++ b/.story_kit/work/1_upcoming/230_story_prevent_duplicate_stage_agents_on_same_story.md
@@ -0,0 +1,22 @@
+---
+name: "Prevent duplicate stage agents on same story"
+---
+
+# Story 230: Prevent duplicate stage agents on same story
+
+## User Story
+
+As a supervisor, I want start_agent to reject a second coder (or QA, or mergemaster) on a story that already has one running, so that two agents don't operate in the same worktree and corrupt each other's work.
+
+## Acceptance Criteria
+
+- [ ] start_agent rejects a second coder-stage agent on a story that already has a Running or Pending coder, with a clear error message naming both agents
+- [ ] The check covers all non-Other pipeline stages (Coder, Qa, Mergemaster), not just Coder
+- [ ] The check fires for both explicit agent_name requests and auto-selected agents
+- [ ] Auto-assign (auto_assign_available_work) is unaffected — it already has its own is_story_assigned_for_stage guard
+- [ ] Regression test: concurrent start_agent calls with two different coder names on the same story — exactly one succeeds
+- [ ] Regression test: two coders on different stories still works fine
+
+## Out of Scope
+
+- TBD