diff --git a/.graphlemon/runs/3605272e-fe33-48cc-b3f3-e0c0ba26f970/benchmark.md b/.graphlemon/runs/3605272e-fe33-48cc-b3f3-e0c0ba26f970/benchmark.md
new file mode 100644
index 0000000..cb82780
--- /dev/null
+++ b/.graphlemon/runs/3605272e-fe33-48cc-b3f3-e0c0ba26f970/benchmark.md
@@ -0,0 +1,9 @@
+The graphlemon Work Node (planner) workflow finished. It cross-checked the SDD against the current codebase (`frontend/app.js`, `src/work_graph_run_query.py`, `src/task_query.py`, `src/backend_server.py`, `src/schema.py`, and related tests) and produced a Benchmark artifact for **U-046: "Run history click navigates incorrectly + missing run detail page"**, covering:
+
+- **T1** — new `get_work_graph_run(conn, run_id)` query function contract (happy/boundary/not-found cases)
+- **T2** — new `GET /api/work-graph-runs/<run_id>` route (must not collide with the existing list route, 404 conventions matching other single-resource GETs)
+- **T3** — frontend: change the `gc_runs_list` row click to navigate via `data-run-id` + hash change instead of `onSwitchToTask`, without altering existing list rendering
+- **T4** — new `#/runs/:run_id` route and run-detail page, reusing `renderGraphVisualizerPanel`/`mountLogViewer`/route-cleanup patterns, plus edge cases (empty DAG, non-existent run, WS cleanup on navigation)
+- **T5** — regression: full `src` unittest suite and existing frontend smoke tests must stay green
+
+Full benchmark text is saved in the task output file if the complete document is needed. Nothing was committed — this was planning/analysis only. Ready to proceed to implementation (Coder) or hand off to the next Work Node stage if desired.
\ No newline at end of file