Закрепить целевой AGENT-прогон hot value-flow handoff

This commit is contained in:
2026-05-22 23:00:37 +03:00
parent e7603a9d29
commit 50d938b8f1
14 changed files with 636 additions and 8 deletions
@@ -0,0 +1,157 @@
{
"schema_version": "domain_truth_harness_spec_v1",
"scenario_id": "agent_hot_value_flow_handoff_20260522",
"domain": "autonomy_hot_value_flow_handoff",
"title": "AGENT | Hot value-flow discovery handoff",
"description": "Targeted AGENT replay for the current Autonomy Core slice: organization-scoped value-flow questions must be answered through guarded MCP discovery response with hot handoff, not through stale exact fallback or a counterparty-only route.",
"bindings": {},
"steps": [
{
"step_id": "step_01_incoming_total_hot_handoff",
"title": "Organization-scoped incoming total uses hot value-flow discovery candidate",
"question": "Сколько входящих денег за 2020 год по ООО Альтернатива Плюс без разреза по контрагентам?",
"allowed_reply_types": [
"partial_coverage",
"factual_with_explanation",
"factual"
],
"expected_mcp_discovery_response_applied": true,
"expected_mcp_discovery_selected_chain_id": "value_flow",
"expected_mcp_discovery_response_candidate_status": "ready_for_guarded_use",
"expected_mcp_discovery_candidate_hot_runtime_wired": true,
"expected_mcp_discovery_hot_runtime_wired": true,
"expected_mcp_discovery_execution_handoff_status": "ready_for_guarded_response",
"expected_mcp_discovery_execution_handoff_can_use_guarded_response": true,
"expected_catalog_alignment_status": "selected_matches_top",
"expected_catalog_chain_top_match": "value_flow",
"expected_catalog_selected_matches_top": true,
"expected_route_candidate_status": "ready_for_reviewed_execution",
"expected_route_candidate_executable_now": true,
"required_answer_patterns_all": [
"(?i)2020",
"(?i)входящ|получ|поступ",
"(?i)руб"
],
"required_answer_patterns_any": [
"(?i)Альтернатива",
"(?i)проверенн",
"(?i)1С"
],
"forbidden_answer_patterns": [
"(?i)уточните контрагента",
"(?i)по какому контрагенту",
"(?i)не найден контрагент",
"(?i)runtime_",
"(?i)planner_",
"(?i)query_movements",
"(?i)primitive"
],
"criticality": "critical",
"semantic_tags": [
"autonomy_core",
"value_flow",
"hot_handoff",
"guarded_response",
"incoming_total"
]
},
{
"step_id": "step_02_outgoing_total_hot_handoff",
"title": "Organization-scoped outgoing total uses hot value-flow discovery candidate",
"question": "Сколько исходящих денег за 2020 год по ООО Альтернатива Плюс без разреза по контрагентам?",
"allowed_reply_types": [
"partial_coverage",
"factual_with_explanation",
"factual"
],
"expected_mcp_discovery_response_applied": true,
"expected_mcp_discovery_selected_chain_id": "value_flow",
"expected_mcp_discovery_response_candidate_status": "ready_for_guarded_use",
"expected_mcp_discovery_candidate_hot_runtime_wired": true,
"expected_mcp_discovery_hot_runtime_wired": true,
"expected_mcp_discovery_execution_handoff_status": "ready_for_guarded_response",
"expected_mcp_discovery_execution_handoff_can_use_guarded_response": true,
"expected_catalog_alignment_status": "selected_matches_top",
"expected_catalog_chain_top_match": "value_flow",
"expected_catalog_selected_matches_top": true,
"expected_route_candidate_status": "ready_for_reviewed_execution",
"expected_route_candidate_executable_now": true,
"required_answer_patterns_all": [
"(?i)2020",
"(?i)исходящ|списан|заплат",
"(?i)руб"
],
"required_answer_patterns_any": [
"(?i)Альтернатива",
"(?i)проверенн",
"(?i)1С"
],
"forbidden_answer_patterns": [
"(?i)уточните контрагента",
"(?i)по какому контрагенту",
"(?i)не найден контрагент",
"(?i)runtime_",
"(?i)planner_",
"(?i)query_movements",
"(?i)primitive"
],
"criticality": "critical",
"semantic_tags": [
"autonomy_core",
"value_flow",
"hot_handoff",
"guarded_response",
"outgoing_total"
]
},
{
"step_id": "step_03_colloquial_money_total_hot_handoff",
"title": "Colloquial company money wording still uses hot value-flow discovery candidate",
"question": "А всего сколько денег пришло в ООО Альтернатива Плюс за 2020, без топов и без контрагентов?",
"allowed_reply_types": [
"partial_coverage",
"factual_with_explanation",
"factual"
],
"expected_mcp_discovery_response_applied": true,
"expected_mcp_discovery_selected_chain_id": "value_flow",
"expected_mcp_discovery_response_candidate_status": "ready_for_guarded_use",
"expected_mcp_discovery_candidate_hot_runtime_wired": true,
"expected_mcp_discovery_hot_runtime_wired": true,
"expected_mcp_discovery_execution_handoff_status": "ready_for_guarded_response",
"expected_mcp_discovery_execution_handoff_can_use_guarded_response": true,
"expected_catalog_alignment_status": "selected_matches_top",
"expected_catalog_chain_top_match": "value_flow",
"expected_catalog_selected_matches_top": true,
"expected_route_candidate_status": "ready_for_reviewed_execution",
"expected_route_candidate_executable_now": true,
"required_answer_patterns_all": [
"(?i)2020",
"(?i)пришл|получ|поступ|входящ",
"(?i)руб"
],
"required_answer_patterns_any": [
"(?i)Альтернатива",
"(?i)проверенн",
"(?i)1С"
],
"forbidden_answer_patterns": [
"(?i)уточните контрагента",
"(?i)по какому контрагенту",
"(?i)не найден контрагент",
"(?i)runtime_",
"(?i)planner_",
"(?i)query_movements",
"(?i)primitive"
],
"criticality": "critical",
"semantic_tags": [
"autonomy_core",
"value_flow",
"hot_handoff",
"guarded_response",
"colloquial_total"
]
}
]
}