diff --git a/CODESMELL.md b/CODESMELL.md
index 4bd423f..53a456d 100644
--- a/CODESMELL.md
+++ b/CODESMELL.md
@@ -72,18 +72,17 @@
 
 ---
 
-### CSM-004 — Blocking `init/1` + Missing `terminate/2` in Job Runner
-- **File:** `lib/bds/scripting/job_runner.ex:28-35`
-- **What:**
-  1. Synchronous `GenServer.call` to `JobStore.attach_runner/2` inside `init/1` (Z. 30) blocks the supervisor if `JobStore` is slow.
-  2. `Process.flag(:trap_exit, true)` (Z. 29) is set but there is **no** `terminate/2` callback and **no** `{:EXIT, ...}` handler in `handle_info`.
-- **Why it's bad:**
-  - Supervisor child startup hangs if `JobStore` is slow or unavailable.
-  - If the runner process is killed (not via `:cancel`), the PID leaks in `JobStore.runners`. The `detach_runner/2` call only happens in `handle_call(:cancel)` (Z. 71) and `handle_info({ref, result})` (Z. 95) and `handle_info({:DOWN, ...})` (Z. 117). A linked process crash with `trap_exit` would send `{:EXIT, pid, reason}` which has **no handler**, so the runner stops without detaching.
-- **Fix:**
-  - Move `attach_runner` to `handle_continue/2`.
-  - Either remove `trap_exit` (the task uses `async_nolink`, so links aren't the issue) OR add `terminate/2` that calls `JobStore.detach_runner/2` and add an `{:EXIT, ...}` clause in `handle_info`.
-- **Test:** Start a runner, kill it (not via cancel); assert `JobStore` does not contain the dead PID.
+### ~~CSM-004 — Blocking `init/1` + Missing `terminate/2` in Job Runner~~ ✅ FIXED
+- **Fixed:** 2026-05-08
+- **What was done:**
+  - Moved `JobStore.attach_runner/2` from `init/1` to a new `handle_continue(:attach_and_start)` callback, so supervisor startup is no longer blocked by the synchronous call.
+  - Added `terminate/2` callback that calls `JobStore.detach_runner/2` (with `try/catch` for shutdown safety), centralizing cleanup that was previously scattered across individual exit paths.
+  - Added `handle_info({:EXIT, _pid, _reason})` clause to handle trapped exit signals from linked processes.
+  - Removed redundant inline `detach_runner` calls from `handle_call(:cancel)`, task result handler, and `:DOWN` handler — `terminate/2` now handles all detach cleanup.
+  - Changed `restart: :temporary` since job runners are one-shot processes that should not auto-restart on failure.
+  - Added `@impl true` to all `handle_info` clauses.
+  - Fixed pre-existing bug in `JobStore.detach_runner` handler where `update_in/2` macro result was incorrectly double-wrapped, corrupting state.
+  - Added test: start a runner, kill it externally (not via cancel), assert `JobStore` no longer contains the dead PID.
 
 ---
 
@@ -400,6 +399,7 @@
 - [x] All critical items (CSM-001 to CSM-005) have been addressed or explicitly deferred with justification.
   - CSM-001: Fixed. All `String.to_atom` on dynamic data replaced with `MapUtils.safe_atomize_key/keys` or `String.to_existing_atom`.
   - CSM-002: Fixed. Search now pushes all filtering and pagination into SQL via Ecto queries and CTEs.
+  - CSM-004: Fixed. `attach_runner` moved to `handle_continue`, `terminate/2` added for cleanup, `restart: :temporary` set, JobStore `detach_runner` bug fixed.
 - [ ] All high-severity items (CSM-006 to CSM-010) have been addressed.
 - [x] CSM-001 fix covers ALL 6 affected files, not just `import_definitions.ex`.
 - [x] CSM-003 fix covers ALL `Repo.delete!` call sites (posts, tags, scripts, media, projects, templates, translations).
diff --git a/lib/bds/scripting/job_runner.ex b/lib/bds/scripting/job_runner.ex
index 1d67c69..b28c061 100644
--- a/lib/bds/scripting/job_runner.ex
+++ b/lib/bds/scripting/job_runner.ex
@@ -1,7 +1,7 @@
 defmodule BDS.Scripting.JobRunner do
   @moduledoc false
 
-  use GenServer
+  use GenServer, restart: :temporary
 
   def start_link(opts) do
     GenServer.start_link(__MODULE__, opts)
@@ -27,8 +27,13 @@ defmodule BDS.Scripting.JobRunner do
     }
 
     Process.flag(:trap_exit, true)
+    {:ok, state, {:continue, :attach_and_start}}
+  end
+
+  @impl true
+  def handle_continue(:attach_and_start, state) do
     :ok = BDS.Scripting.JobStore.attach_runner(state.job_id, self())
-    {:ok, state, {:continue, :start_job}}
+    {:noreply, state, {:continue, :start_job}}
   end
 
   @impl true
@@ -68,7 +73,6 @@ defmodule BDS.Scripting.JobRunner do
         finished_at: DateTime.utc_now()
       })
 
-    :ok = BDS.Scripting.JobStore.detach_runner(state.job_id)
     {:stop, :normal, :ok, %{state | cancelled?: true}}
   end
 
@@ -78,6 +82,7 @@ defmodule BDS.Scripting.JobRunner do
     {:noreply, state}
   end
 
+  @impl true
   def handle_info({ref, result}, %{task_ref: ref} = state) do
     Process.demonitor(ref, [:flush])
 
@@ -92,12 +97,12 @@ defmodule BDS.Scripting.JobRunner do
         end
 
       :ok = BDS.Scripting.JobStore.update_job(state.job_id, attrs)
-      :ok = BDS.Scripting.JobStore.detach_runner(state.job_id)
     end
 
     {:stop, :normal, %{state | completed?: true}}
   end
 
+  @impl true
   def handle_info({:DOWN, ref, :process, _pid, reason}, %{task_ref: ref} = state) do
     cond do
       state.completed? or state.cancelled? ->
@@ -114,8 +119,21 @@ defmodule BDS.Scripting.JobRunner do
             finished_at: DateTime.utc_now()
           })
 
-        :ok = BDS.Scripting.JobStore.detach_runner(state.job_id)
         {:stop, :normal, state}
     end
   end
+
+  @impl true
+  def handle_info({:EXIT, _pid, _reason}, state) do
+    {:stop, :normal, state}
+  end
+
+  @impl true
+  def terminate(_reason, state) do
+    try do
+      BDS.Scripting.JobStore.detach_runner(state.job_id)
+    catch
+      :exit, _ -> :ok
+    end
+  end
 end
diff --git a/lib/bds/scripting/job_store.ex b/lib/bds/scripting/job_store.ex
index c83ea39..bd0a38a 100644
--- a/lib/bds/scripting/job_store.ex
+++ b/lib/bds/scripting/job_store.ex
@@ -65,8 +65,7 @@ defmodule BDS.Scripting.JobStore do
   end
 
   def handle_call({:detach_runner, job_id}, _from, state) do
-    next_state = update_in(state.runners, &Map.delete(&1, job_id))
-    {:reply, :ok, %{state | runners: next_state}}
+    {:reply, :ok, %{state | runners: Map.delete(state.runners, job_id)}}
   end
 
   def handle_call({:fetch_job, job_id}, _from, state) do
diff --git a/test/bds/scripting/job_test.exs b/test/bds/scripting/job_test.exs
index 183b6f7..aa97a9e 100644
--- a/test/bds/scripting/job_test.exs
+++ b/test/bds/scripting/job_test.exs
@@ -90,6 +90,26 @@ defmodule BDS.Scripting.JobTest do
     assert cancelled_job.finished_at != nil
   end
 
+  test "killing a runner detaches it from JobStore (CSM-004)" do
+    Application.put_env(:bds, :scripting,
+      runtime: BlockingRuntime,
+      timeout: 300_000,
+      max_reductions: 5_000_000,
+      job_timeout: :infinity,
+      job_max_reductions: :none
+    )
+
+    assert {:ok, job} = BDS.Scripting.start_job("irrelevant", "main")
+    _running_job = wait_for_job(job.id, &(&1.status == :running))
+
+    runner_pid = BDS.Scripting.JobStore.runner_for(job.id)
+    assert is_pid(runner_pid)
+
+    GenServer.stop(runner_pid, :shutdown)
+    Process.sleep(50)
+    assert BDS.Scripting.JobStore.runner_for(job.id) == nil
+  end
+
   defp wait_for_job(job_id, predicate, attempts \\ 50)
 
   defp wait_for_job(job_id, predicate, attempts) when attempts > 0 do