diff --git a/agentlab/reports/capability_digest.py b/agentlab/reports/capability_digest.py
index d132ce1..ed6555c 100644
--- a/agentlab/reports/capability_digest.py
+++ b/agentlab/reports/capability_digest.py
@@ -171,6 +171,7 @@ def _run_context_lines(context: MarkdownRunContext) -> List[str]:
                 "Cost USD",
                 "Duration ms",
                 "Scope Oracle",
+                "Task Provenance",
             ],
             [_trial_row(result) for result in context.results],
         )
@@ -450,9 +451,23 @@ def _trial_row(result: OutcomeEvidence) -> List[object]:
         _unknown_if_none(result.cost_usd),
         result.duration_ms,
         compact_scope_oracle_metadata(result.scope_oracle),
+        _trial_provenance(result),
     ]
 
 
+def _trial_provenance(result: OutcomeEvidence) -> str:
+    return "; ".join(
+        [
+            f"repo={_display_unknown(result.task_repo)}",
+            f"commit={_display_unknown(result.task_commit)}",
+            (
+                "workspace_base_ref="
+                f"{_format_run_surface_value(result.run_surface.get('workspace_base_ref'))}"
+            ),
+        ]
+    )
+
+
 def _markdown_table(headers: List[str], rows: List[List[object]]) -> List[str]:
     table = [
         "| " + " | ".join(_escape_cell(header) for header in headers) + " |",
@@ -508,6 +523,11 @@ def _unknown_if_none(value: object) -> object:
     return value
 
 
+def _display_unknown(value: object) -> str:
+    text = "" if value is None else str(value)
+    return text or "unknown"
+
+
 def _escape_cell(value: object) -> str:
     text = "" if value is None else str(value)
     return text.replace("|", "\\|").replace("\n", "<br>")
diff --git a/agentlab/reports/capability_digest_html.py b/agentlab/reports/capability_digest_html.py
index 72a41ae..36a3859 100644
--- a/agentlab/reports/capability_digest_html.py
+++ b/agentlab/reports/capability_digest_html.py
@@ -69,6 +69,7 @@
     "Secondary Review Labels",
     "Primary Review Label",
     "Exclusions",
+    "Task Provenance",
     "Scope Oracle",
     "Operability Dimension",
     "Evidence",
@@ -500,7 +501,7 @@ def _context_page(
     {_summary_section("Outcome Summary", _outcome_rows(context, render_context), ["Task", "Type", "Total", "Fair", "Excluded", "Passes", "Accepted", "Pass Rate", "pass@k", "pass^k"])}
     {_summary_section("Token Summary", _token_rows(context, render_context), ["Task", "Type", "IO Tokens", "Cached Tokens", "Reason Tokens", "IO Tok / Verified", "IO Tok / Accepted", "Cached Tok / Verified", "Reason Tok / Verified"])}
     {_summary_section("Review and Patch Summary", _review_rows(context, render_context), ["Task", "Type", "Median ms", "Median Files", "Median +Lines", "Median -Lines", "Primary Review Labels", "Secondary Review Labels", "Exclusions"], note=_review_patch_size_caveat_note(context))}
-    {_summary_section("Trial Evidence", _trial_rows(context, render_context), ["Task", "Type", "Trial", "Grader Outcome", "Validity", "Primary Review Label", "Secondary Review Labels", "Exclusion", "Files", "+Lines", "-Lines", "Input Tokens", "Cached Input Tokens", "Output Tokens", "Reasoning Tokens", "Cost USD", "Duration ms", "Scope Oracle", "Report", "Transcript", "Diff", "Result"], note=_trial_evidence_note(context.results))}
+    {_summary_section("Trial Evidence", _trial_rows(context, render_context), ["Task", "Type", "Trial", "Grader Outcome", "Validity", "Primary Review Label", "Secondary Review Labels", "Exclusion", "Files", "+Lines", "-Lines", "Input Tokens", "Cached Input Tokens", "Output Tokens", "Reasoning Tokens", "Cost USD", "Duration ms", "Scope Oracle", "Task Provenance", "Report", "Transcript", "Diff", "Result"], note=_trial_evidence_note(context.results))}
   </main>
 """
 
@@ -721,6 +722,7 @@ def _trial_rows(
             "Task": _task_link_from_result(result, render_context),
             "Type": _text(result.eval_type),
             "Trial": _text(result.trial_id),
+            "Task Provenance": _trial_provenance(result),
             "Grader Outcome": _text(result.status),
             "Validity": _text(result.trial_validity),
             "Primary Review Label": _text(result.primary_review_label),
@@ -755,6 +757,19 @@ def _trial_rows(
     ]
 
 
+def _trial_provenance(result: OutcomeEvidence) -> str:
+    return "<br>".join(
+        [
+            _text(f"repo={_display_unknown(result.task_repo)}"),
+            _text(f"commit={_display_unknown(result.task_commit)}"),
+            _text(
+                "workspace_base_ref="
+                f"{_format_run_surface_value(result.run_surface.get('workspace_base_ref'))}"
+            ),
+        ]
+    )
+
+
 def _patch_size_caveat_note(results: Sequence[OutcomeEvidence]) -> str:
     if not has_patch_size_caveats(results):
         return ""
@@ -1072,6 +1087,11 @@ def _unknown_if_none(value: object) -> object:
     return value
 
 
+def _display_unknown(value: object) -> str:
+    text = "" if value is None else str(value)
+    return text or "unknown"
+
+
 def _pill(text: str, kind: str = "") -> str:
     return f'<span class="pill {kind}">{_text(text)}</span>'
 
diff --git a/tests/test_evidence.py b/tests/test_evidence.py
index bf4f148..7fd1ab8 100644
--- a/tests/test_evidence.py
+++ b/tests/test_evidence.py
@@ -199,6 +199,95 @@ def test_digest_renders_scope_oracle_metadata(self):
         self.assertIn("allowed_paths=src/", html_report)
         self.assertIn("forbidden_paths=src/private/", html_report)
 
+    def test_digest_maps_trial_provenance_for_mixed_workspace_bases(self):
+        results = [
+            normalize_outcome_evidence(
+                {
+                    "trial_id": "trial-a",
+                    "task_id": "task-a",
+                    "task_repo": "https://github.com/example/repo-a",
+                    "task_commit": "aaa111",
+                    "eval_suite": "starter",
+                    "eval_type": "regression",
+                    "agent_name": "codex",
+                    "model_name": "model-a",
+                    "agent_harness_config": {"reasoning_effort": "xhigh"},
+                    "run_surface": {"workspace_base_ref": "refs/base-a"},
+                    "status": "passed",
+                    "success": True,
+                    "duration_ms": 100,
+                    "files_changed": ["app.py"],
+                    "lines_added": 5,
+                    "lines_deleted": 1,
+                    "run_dir": "runs/trial-a",
+                }
+            ),
+            normalize_outcome_evidence(
+                {
+                    "trial_id": "trial-b",
+                    "task_id": "task-b",
+                    "task_repo": "https://github.com/example/repo-b",
+                    "task_commit": "bbb222",
+                    "eval_suite": "starter",
+                    "eval_type": "regression",
+                    "agent_name": "codex",
+                    "model_name": "model-a",
+                    "agent_harness_config": {"reasoning_effort": "xhigh"},
+                    "run_surface": {"workspace_base_ref": "refs/base-b"},
+                    "status": "failed",
+                    "success": False,
+                    "duration_ms": 200,
+                    "files_changed": ["app.py"],
+                    "lines_added": 3,
+                    "lines_deleted": 0,
+                    "run_dir": "runs/trial-b",
+                }
+            ),
+        ]
+
+        digest = render_capability_evidence_digest(results)
+        html_report = render_capability_evidence_digest_html(results)
+
+        self.assertIn("Workspace Base Ref", digest)
+        self.assertIn("mixed: refs/base-a; refs/base-b", digest)
+        self.assertIn(
+            "| Task | Type | Trial | Grader Outcome | Validity |",
+            digest,
+        )
+        self.assertIn("Task Provenance", digest)
+        self.assertRegex(
+            digest,
+            (
+                r"\| task-a \| regression \| trial-a \| passed \| valid \|.*"
+                r"repo=https://github\.com/example/repo-a; "
+                r"commit=aaa111; workspace_base_ref=refs/base-a \|"
+            ),
+        )
+        self.assertRegex(
+            digest,
+            (
+                r"\| task-b \| regression \| trial-b \| failed \| valid \|.*"
+                r"repo=https://github\.com/example/repo-b; "
+                r"commit=bbb222; workspace_base_ref=refs/base-b \|"
+            ),
+        )
+        self.assertIn("Task Provenance", html_report)
+        self.assertIn("mixed: refs/base-a; refs/base-b", html_report)
+        self.assertIn(
+            (
+                "repo=https://github.com/example/repo-a<br>"
+                "commit=aaa111<br>workspace_base_ref=refs/base-a"
+            ),
+            html_report,
+        )
+        self.assertIn(
+            (
+                "repo=https://github.com/example/repo-b<br>"
+                "commit=bbb222<br>workspace_base_ref=refs/base-b"
+            ),
+            html_report,
+        )
+
     def test_nested_setup_created_untracked_patch_size_caveat_round_trips(self):
         result = normalize_outcome_evidence(
             {