Pass --output-format json with --json-schema and unwrap CLI envelope

The Claude Code adapter previously passed --json-schema alone. On Claude
CLI 2.1.160 that combination still emits the model's conversational
preamble on stdout while the schema-enforced structured payload ships on
a sidecar channel the adapter cannot read. Result: callers requesting
structured output got prose that fails JSON parsing downstream — exactly
the failure mode the activity-core CUST-WP-0045 daily triage canary hit
on 2026-06-01 ("Triage report generated and returned via structured
output. Key signals:..." → json.loads error at column 1).

Fix: when --json-schema is set, also pass --output-format json. The CLI
then writes a JSON envelope on stdout. The adapter unwraps it by
probing a small allowlist of known text-bearing fields (result,
result_text, content, text, output). Unknown envelope shapes fall
through to raw stdout so the operator can introspect the structure and
extend the allowlist.

The unwrap path is only triggered when --json-schema was set, so non-
schema callers keep the existing raw-stdout behavior.

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

tests/test_claude_code.py

@@ -39,15 +39,65 @@ def test_execute_prompt_passes_json_schema_to_claude_cli(monkeypatch):
             "--print",
             "--json-schema",
             '{"type":"object"}',
+            "--output-format",
+            "json",
         ],
         "input": "Produce a report.",
         "capture_output": True,
         "text": True,
         "timeout": 42,
     }
+    # Stdout shape that does not match any known envelope field is returned
+    # verbatim so the caller can introspect and we can extend the field list.
     assert response.content == '{"summary":"ok","recommendations":[]}'
 
 
+def test_execute_prompt_unwraps_cli_json_envelope_result_field(monkeypatch):
+    """With --output-format json the CLI wraps the model payload in an
+    envelope. The adapter unwraps the textual result so the caller still
+    sees the model's structured-output JSON, not the envelope."""
+    def fake_run(cmd, input, capture_output, text, timeout):  # noqa: ANN001
+        envelope = {
+            "type": "result",
+            "result": '{"summary":"ok","recommendations":[]}',
+            "total_cost_usd": 0.001,
+        }
+        import json as _json
+        return SimpleNamespace(
+            returncode=0,
+            stdout=_json.dumps(envelope),
+            stderr="",
+        )
+
+    monkeypatch.setattr("llm_connect.claude_code.subprocess.run", fake_run)
+    adapter = ClaudeCodeAdapter(cli_path="/custom/claude")
+
+    response = adapter.execute_prompt(
+        "Produce a report.",
+        RunConfig(model_params={"json_schema": {"type": "object"}}),
+    )
+
+    assert response.content == '{"summary":"ok","recommendations":[]}'
+
+
+def test_execute_prompt_no_unwrap_without_json_schema(monkeypatch):
+    """Without --json-schema we do not pass --output-format json, so the
+    envelope unwrap path stays inert and raw stdout passes through."""
+    def fake_run(cmd, input, capture_output, text, timeout):  # noqa: ANN001
+        return SimpleNamespace(
+            returncode=0,
+            stdout='{"result":"this is just stdout, not an envelope"}',
+            stderr="",
+        )
+
+    monkeypatch.setattr("llm_connect.claude_code.subprocess.run", fake_run)
+    adapter = ClaudeCodeAdapter(cli_path="/custom/claude")
+
+    response = adapter.execute_prompt("Plain prompt.", RunConfig())
+
+    assert response.content == '{"result":"this is just stdout, not an envelope"}'
+
+
 def test_claude_code_adapter_prefers_env_cli_path(monkeypatch):
     monkeypatch.setenv("LLM_CONNECT_CLAUDE_CLI_PATH", "/home/me/bin/claude")