diff --git a/.huskies/work/1_backlog/489_story_sqlite_shadow_write_for_pipeline_state_via_sqlx.md b/.huskies/work/1_backlog/489_story_sqlite_shadow_write_for_pipeline_state_via_sqlx.md
new file mode 100644
index 00000000..da61dec0
--- /dev/null
+++ b/.huskies/work/1_backlog/489_story_sqlite_shadow_write_for_pipeline_state_via_sqlx.md
@@ -0,0 +1,25 @@
+---
+name: "SQLite shadow write for pipeline state via sqlx"
+---
+
+# Story 489: SQLite shadow write for pipeline state via sqlx
+
+## User Story
+
+As a developer, I want pipeline state dual-written to SQLite (via sqlx) alongside the existing filesystem directories, so we have a database layer ready for CRDT integration without changing any existing behaviour.
+
+## Acceptance Criteria
+
+- [ ] Add sqlx with SQLite feature as a dependency
+- [ ] Migrations embedded at compile time via sqlx::migrate! macro, run on startup
+- [ ] Schema uses backend-agnostic SQL (TEXT, INTEGER, no vendor-specific types) so migrations work on both SQLite and Postgres
+- [ ] Every move_story_to_X and pipeline state change writes to both .huskies/work/ directories AND SQLite
+- [ ] Reads still come from the filesystem (SQLite is shadow-only)
+- [ ] SQLite database stored at .huskies/pipeline.db
+- [ ] pipeline_items table: id, name, stage, agent, retry_count, blocked, depends_on, created_at, updated_at
+- [ ] All existing pipeline operations work unchanged from the user's perspective
+- [ ] agent: coder-opus
+
+## Out of Scope
+
+- TBD