fix(explore): preserve current-task evidence order across planner links

loopx/capabilities/explore/README.md

@@ -694,6 +694,13 @@ or quota, and neither mode is enabled by supplying a task id.
 The dashboard's saved-evidence reader also pages through history; the CLI
 options above control the agent context, not dashboard page size.
 
+Within the current task's linked questions, results retain source recency after
+the reserved counterexamples; measurement groups retain their own append-order
+recency. A question shared with a different planned task gets no extra priority.
+Planner ordering applies only outside the current task's links. When diagnosing
+buried evidence, compare the ordinary task page with the node-filtered read:
+successful lookup alone does not show that the next agent saw or used it.
+
 When the stride strategy is enabled, findings read from a global result page
 can be cited in `explore_review.considered_refs` by also copying that page's
 `revision`, `offset` and `limit` into `explore_review.result_pages` (up to three

loopx/capabilities/explore/README.zh-CN.md

@@ -211,6 +211,11 @@ Todo 关联保留既有可恢复边界与 claim/lease 证明；恢复旧批次
 每个相关问题仍优先保留一条范围反证；其它结果继续分页可读，缺少关联不代表没有证据。
 该读取不启用能力、不修改规划分数、任务准入、claim/lease、配额或图谱。
 
+当前任务关联的问题内部，结果在保留范围反证后按原来源新旧排序，量化组保持自身的
+追加顺序；某个问题同时关联另一条计划任务，不会因此获得额外优先级。规划器排序只
+适用于当前任务关联之外的结果。排查证据被埋没时，应比较普通任务页与按问题筛选的
+读取；查询成功本身不证明下一轮 agent 已看到或采用该证据。
+
 任务范围的 `result_page` 多携带 `todo_id`，后续命令也保留它。引用该页时，将它和
 `revision/offset/limit` 一同复制到原 `explore_review.result_pages`；主 review revision
 仍引用原有全局 basis。写回从同一可信来源重放任务关联与精确页面，不替换已准入的

loopx/control_plane/capabilities/explore_turn_context.ts

@@ -61,7 +61,7 @@ function branchContext(row: JsonObject, findingSources?: Map<string, JsonObject>
 }
 /** Keep the latest scoped counterexample per linked question before its other
  * findings; recency alone must not supersede evidence on different inputs. */
-function prioritizeLinkedResults(results: JsonObject[], refs: Set<string>): JsonObject[] {
+function prioritizeLinkedResults(results: JsonObject[], refs: Set<string>, fallbackResults = results): JsonObject[] {
   const linked = results.filter(row => refs.has(String(row.node_id ?? "")));
   const refuted = new Set<string>();
   const counterexamples = linked.filter(row => {
@@ -72,7 +72,7 @@ function prioritizeLinkedResults(results: JsonObject[], refs: Set<string>): Json
   });
   const reserved = new Set(counterexamples);
   return [...counterexamples, ...linked.filter(row => !reserved.has(row)),
-    ...results.filter(row => !refs.has(String(row.node_id ?? "")))];
+    ...fallbackResults.filter(row => !refs.has(String(row.node_id ?? "")))];
 }
 export function projectExploreTurnContext(params: JsonObject): JsonObject {
   const goal = requireNonEmptyString(params.goal_id, "goal_id");
@@ -118,7 +118,10 @@ export function projectExploreTurnContext(params: JsonObject): JsonObject {
   // Explicitly bound work can differ from this independent planner. Prioritize
   // its canonical links without changing candidates, scores or admission.
   const plannerResults = prioritizeLinkedResults(attachedResults, selectedRefs);
-  const writebackResults = focusedRefs.size ? prioritizeLinkedResults(plannerResults, focusedRefs) : plannerResults;
+  // Shared links must not inherit another task's planner priority. Use source
+  // recency within the focus and retain planner ordering only outside it.
+  const writebackResults = focusedRefs.size
+    ? prioritizeLinkedResults(attachedResults, focusedRefs, plannerResults) : plannerResults;
   const command = (...args: string[]) => [...route, "explore", ...args, "--goal-id", goal];
   const resultLimit = requireInteger(params.result_limit ?? 3, "result_limit");
   const resultOffset = requireInteger(params.result_offset ?? 0, "result_offset");
@@ -129,7 +132,8 @@ export function projectExploreTurnContext(params: JsonObject): JsonObject {
   const measurementFeedback = projectExploreMeasurementFeedback(findings)
     .filter(item => resultNode === null || item.node_id === resultNode)
     .sort((a, b) => Number(focusedRefs.has(b.node_id)) - Number(focusedRefs.has(a.node_id))
-      || Number(selectedRefs.has(b.node_id)) - Number(selectedRefs.has(a.node_id)));
+      || (focusedRefs.has(a.node_id) ? 0
+        : Number(selectedRefs.has(b.node_id)) - Number(selectedRefs.has(a.node_id))));
   const resultRows = resultNode == null ? writebackResults : writebackResults.filter(row => row.node_id === resultNode);
   const nodeById = new Map(nodes.map(node => [node.node_id, node]));
   // Bind current Todo links to the results on this page. Include the full

tests/control_plane/test_explore_task_scope.py

@@ -85,3 +85,54 @@ def cli(*args, expected=0):
     stale = cli(*next_args[next_args.index("explore"):], expected=1)
     assert "changed" in stale["error"] and log.read_bytes() == source
     cli("explore", "turn-context", *identity, "--todo-id", "todo_missing", expected=1)
+
+
+@pytest.mark.parametrize("provider", ["file", "sqlite"])
+def test_shared_planner_question_does_not_bury_new_current_task_evidence(tmp_path, monkeypatch, provider):
+    isolate_sqlite_runtime(tmp_path, monkeypatch)
+    monkeypatch.setenv("LOOPX_USAGE_PING", "0")
+    project, runtime, registry = fixture._write_fixture(tmp_path)
+    monkeypatch.chdir(project)
+    fixture._initialize_fixture_authority(project, runtime, provider=provider)
+
+    def cli(*args):
+        code, packet = fixture._run_cli(registry, runtime, *args, cwd=project)
+        assert code == 0, json.dumps(packet)
+        return packet
+
+    identity = ("--goal-id", fixture.GOAL_ID, "--agent-id", fixture.AGENT_ID)
+    cli("configure-goal", "--goal-id", fixture.GOAL_ID, "--explore-mode", "planning", "--execute")
+    cli("todo", "update", *identity, "--todo-id", fixture.TODO_ID, "--priority", "P2",
+        "--explore-result-node-ref", "current", "--explore-result-node-ref", "shared")
+    for i in range(4):
+        other = cli("todo", "add", "--goal-id", fixture.GOAL_ID, "--role", "agent", "--priority", "P1",
+            "--claimed-by", fixture.AGENT_ID, "--text", f"Validate separate method {i}.",
+            "--task-class", "advancement_task", "--action-kind", "validate", "--target-key", f"other-{i}")
+        cli("todo", "update", *identity, "--todo-id", other["todo_id"], "--explore-result-node-ref", "shared")
+    log = explore_result_log_path(runtime, fixture.GOAL_ID)
+    for node in ["current", "shared"]:
+        append_explore_result_event(log, build_explore_node_event(
+            goal_id=fixture.GOAL_ID, agent_id=fixture.AGENT_ID, node_id=node,
+            title=node, node_kind="question", status="exploring"))
+    for i, node in enumerate(["shared", "shared", "current"]):
+        evidence = f"result:observation-{i}"
+        append_explore_result_event(log, build_explore_finding_event(
+            goal_id=fixture.GOAL_ID, agent_id=fixture.AGENT_ID, node_id=node,
+            finding_id=f"observation-{i}", title=f"Observation {i}", status="confirmed",
+            summary="Applicable to the named input only.", evidence_refs=[evidence],
+            tags=["writeback-result"], recorded_at=f"2026-01-0{i + 1}T00:00:00Z",
+            measurements=[dict(metric="quality", value=i, unit="points", higher_is_better=True,
+                comparison_scope=f"validation:{node}", input_revision="inputs-v1",
+                evaluation_ref="evaluator:v1", candidate_ref=f"candidate:{i}", evidence_ref=evidence)]))
+    original_log = log.read_bytes()
+    unscoped = cli("explore", "turn-context", *identity, "--result-limit", "1")
+    assert unscoped["graph"]["writeback_results"][0]["node_id"] == "shared"
+    scoped = cli("explore", "turn-context", *identity, "--todo-id", fixture.TODO_ID, "--result-limit", "1")
+    assert scoped["graph"]["writeback_results"][0]["finding_id"] == "observation-2"
+    assert scoped["graph"]["measurement_feedback"][0]["node_id"] == "current"
+    assert scoped["harness"]["selected_branches"] == unscoped["harness"]["selected_branches"]
+    next_args = scoped["graph"]["result_page"]["next_command"]
+    after = cli(*next_args[next_args.index("explore"):])
+    assert after["graph"]["writeback_results"][0]["finding_id"] == "observation-1"
+    assert after["graph"]["measurement_feedback"][0]["node_id"] == "shared"
+    assert log.read_bytes() == original_log

tests/control_plane_ts/explore_turn_context.test.ts

@@ -190,3 +190,43 @@ test("task focus rejects missing, foreign and terminal sources and is inert when
   const off = {...request, graph_enabled: false};
   assert.deepEqual(projectExploreTurnContext(off), projectExploreTurnContext({...off, todo_id: null}));
 });
+
+for (const counterexample of [false, true]) {
+  for (const field of ["writeback_results", "measurement_feedback"]) {
+    test(`overlapping planner links cannot reorder current task ${field}, refutation=${counterexample}`, () => {
+      const finding = (id: string, node: string, status = "confirmed") => ({finding_id: id,
+        node_id: node, status, finding: id, tags: ["writeback-result"],
+        measurements: [{metric: "quality", value: 5, unit: "points", higher_is_better: true,
+          comparison_scope: `validation:${node}`, input_revision: "inputs-v1", evaluation_ref: "eval:v1",
+          candidate_ref: `candidate:${id}`, evidence_ref: `result:${id}`} ]});
+      const findings = [finding("current-new", "current"), finding("shared-old", "shared"),
+        ...(counterexample ? [finding("current-negative", "current", "refuted")] : []),
+        finding("planner-only", "other"), finding("unrelated", "unrelated")];
+      const request = {goal_id: "research", agent_id: "worker", route: ["loopx"], graph_enabled: true,
+        harness_gate: {enabled: true}, result_limit: 1,
+        projection: {nodes: [], findings}, todo_id: "todo_current",
+        focus_todo: {todo_id: "todo_current", role: "agent", status: "open", claimed_by: "worker",
+          explore_result_node_refs: ["current", "shared"]}};
+      const expected = field === "measurement_feedback" ? ["current", "shared", "other", "unrelated"]
+        : [...(counterexample ? ["current-negative"] : []), "current-new", "shared-old", "planner-only", "unrelated"];
+      for (const plannedRefs of [["other"], ["shared", "other"]]) {
+        const params = {...request, plan: {selected_branches: [{todo_id: "todo_planned", actionable_open: true,
+          typed_evidence_audit: {requested_node_refs: plannedRefs}}]}};
+        const first = projectExploreTurnContext(params);
+        const all = projectExploreTurnContext({...params, result_limit: 20});
+        const values = (graph: JsonObject) => graph[field].map((row: JsonObject) =>
+          field === "measurement_feedback" ? row.node_id : row.finding_id);
+        assert.deepEqual(values(first.graph as JsonObject), expected.slice(0, 1));
+        assert.deepEqual(values(all.graph as JsonObject), expected);
+        assert.deepEqual((first.harness as JsonObject).selected_branches,
+          (projectExploreTurnContext({...params, todo_id: null}).harness as JsonObject).selected_branches);
+        if (field === "writeback_results") {
+          const page = {todo_id: request.todo_id, revision: (first.graph as JsonObject).result_page.revision,
+            offset: 1, limit: 1};
+          const replay = projectExploreTurnContext({...params, review_pages: [page]});
+          assert.deepEqual(replay.review_pages[0].finding_refs, expected.slice(1, 2));
+        }
+      }
+    });
+  }
+}