feat(tasks): add skill-templated task graph execution

app-instance/backend/tests/unit/test_task_evidence.py

@@ -3,7 +3,65 @@ from __future__ import annotations
 from pathlib import Path
 
 from beaver.engine.session.manager import SessionManager
-from beaver.tasks.evidence import EvidenceBuilder, RunEvidence, TaskEvidencePacket, ToolEvidence, render_task_evidence
+from beaver.tasks.evidence import (
+    EvidenceBuilder,
+    RunEvidence,
+    TaskEvidencePacket,
+    ToolEvidence,
+    evaluate_node_evidence,
+    render_task_evidence,
+)
+
+
+def _run_evidence(*, tool_results: list[ToolEvidence] | None = None) -> RunEvidence:
+    return RunEvidence(
+        run_id="run-1",
+        session_id="session-1",
+        output_text="",
+        finish_reason="stop",
+        tool_results=list(tool_results or []),
+    )
+
+
+def test_evaluate_node_evidence_requires_successful_tool_result() -> None:
+    evidence = _run_evidence(
+        tool_results=[
+            ToolEvidence(
+                tool_name="web_fetch",
+                tool_call_id="call-1",
+                content="failed",
+                event_payload={"success": False},
+            )
+        ]
+    )
+
+    assert evaluate_node_evidence(evidence, ["tool_result"], "done") == [
+        "missing required evidence: tool_result"
+    ]
+
+
+def test_evaluate_node_evidence_accepts_url_in_successful_tool_content() -> None:
+    evidence = _run_evidence(
+        tool_results=[
+            ToolEvidence(
+                tool_name="web_fetch",
+                tool_call_id="call-1",
+                content="Source: https://example.test/report",
+                event_payload={"success": True},
+            )
+        ]
+    )
+
+    assert evaluate_node_evidence(evidence, ["tool_result", "url"], "done") == []
+
+
+def test_evaluate_node_evidence_checks_output_and_unknown_requirements() -> None:
+    evidence = _run_evidence()
+
+    assert evaluate_node_evidence(evidence, ["output", "unknown_type"], "  ") == [
+        "missing required evidence: output",
+        "unsupported evidence requirement: unknown_type",
+    ]
 
 
 def test_evidence_builder_preserves_full_tool_result(tmp_path: Path) -> None: