diff --git a/.story_kit/work/5_done/317_refactor_split_pool_rs_into_pipeline_auto_assign_and_agent_management_modules.md b/.story_kit/work/5_done/317_refactor_split_pool_rs_into_pipeline_auto_assign_and_agent_management_modules.md
new file mode 100644
index 0000000..79d93f6
--- /dev/null
+++ b/.story_kit/work/5_done/317_refactor_split_pool_rs_into_pipeline_auto_assign_and_agent_management_modules.md
@@ -0,0 +1,29 @@
+---
+name: "Split pool.rs into pipeline, auto-assign, and agent management modules"
+---
+
+# Refactor 317: Split pool.rs into pipeline, auto-assign, and agent management modules
+
+## Current State
+
+- TBD
+
+## Desired State
+
+pool.rs is 5,852 lines and handles agent pool management, pipeline advancement, auto-assign logic, completion handling, and has massive test suites all in one file. Split into focused modules:
+- pool.rs — core agent pool (start, stop, list, inject)
+- pipeline.rs — run_pipeline_advance and stage transition logic
+- auto_assign.rs — auto_assign_available_work and assignment logic
+- Tests split into corresponding test modules
+
+## Acceptance Criteria
+
+- [ ] pool.rs reduced to core agent management (start, stop, list)
+- [ ] Pipeline advancement logic extracted to agents/pipeline.rs
+- [ ] Auto-assign logic extracted to agents/auto_assign.rs
+- [ ] All existing tests pass without modification to test logic
+- [ ] No public API changes — callers still access everything through AgentPool
+
+## Out of Scope
+
+- TBD