{
  "schema_version": "1.0.0",
  "published": "2026-10-09",
  "last_editorial_review": "2026-10-09",
  "provenance_contract": {
    "evidence_origin": [
      "peer_reviewed_external",
      "independent_external",
      "vendor_reported",
      "project_documentation",
      "agentencode_measured"
    ],
    "verification": [
      "reviewed",
      "pending",
      "conflicting",
      "insufficient_evidence"
    ],
    "capability_values": [
      "yes",
      "no",
      "unknown",
      "not_applicable"
    ],
    "unknown_is_false": false,
    "own_benchmarks_executed": false,
    "source_hash_rule": "original_content_sha256 null unless original bytes retrieved, archived and hashed",
    "rights_rule": "No automatic import of third-party result tables, datasets or images without rights review"
  },
  "sources": [
    {
      "source_id": "mas-src-acl-2025",
      "title": "MultiAgentBench: Evaluating the Collaboration and Competition of LLM agents",
      "publisher": "Association for Computational Linguistics",
      "url": "https://aclanthology.org/2025.acl-long.421/",
      "doi": "10.18653/v1/2025.acl-long.421",
      "source_type": "peer_reviewed_paper",
      "published_at": "2025-07",
      "retrieved_at": "2026-10-09",
      "license_status": "ACL Anthology describes CC BY 4.0 for 2016+ materials; component rights not assumed",
      "original_content_sha256": null,
      "hash_status": "original_bytes_not_archived"
    },
    {
      "source_id": "mas-src-marble",
      "title": "MARBLE research repository",
      "publisher": "ulab-uiuc",
      "url": "https://github.com/ulab-uiuc/MARBLE",
      "doi": null,
      "source_type": "official_research_code",
      "published_at": null,
      "retrieved_at": "2026-10-09",
      "license_status": "Repository code MIT; datasets and third-party material require separate review",
      "original_content_sha256": null,
      "hash_status": "original_bytes_not_archived"
    },
    {
      "source_id": "mas-src-agentbench",
      "title": "AgentBench repository",
      "publisher": "THUDM",
      "url": "https://github.com/THUDM/AgentBench",
      "doi": null,
      "source_type": "official_research_code",
      "published_at": "2024",
      "retrieved_at": "2026-10-09",
      "license_status": "not verified for redistribution of full result files",
      "original_content_sha256": null,
      "hash_status": "original_bytes_not_archived"
    },
    {
      "source_id": "mas-src-ms-orch",
      "title": "Agent Framework: workflow orchestrations",
      "publisher": "Microsoft",
      "url": "https://learn.microsoft.com/en-us/agent-framework/workflows/orchestrations/",
      "doi": null,
      "source_type": "project_documentation",
      "published_at": null,
      "retrieved_at": "2026-10-09",
      "license_status": "documentation use reviewed for linking and original paraphrase; bulk import not authorized",
      "original_content_sha256": null,
      "hash_status": "original_bytes_not_archived"
    },
    {
      "source_id": "mas-src-a2a-spec",
      "title": "A2A Protocol Specification",
      "publisher": "A2A Protocol Working Group",
      "url": "https://a2a-protocol.org/latest/specification/",
      "doi": null,
      "source_type": "versioned_specification",
      "published_at": null,
      "retrieved_at": "2026-10-09",
      "specification_version": "1.0",
      "license_status": "no redistribution approval inferred",
      "original_content_sha256": null,
      "hash_status": "original_bytes_not_archived"
    },
    {
      "source_id": "mas-src-mcp-spec",
      "title": "MCP Specification 2025-06-18",
      "publisher": "Model Context Protocol",
      "url": "https://modelcontextprotocol.io/specification/2025-06-18",
      "doi": null,
      "source_type": "versioned_specification",
      "published_at": "2025-06-18",
      "retrieved_at": "2026-10-09",
      "specification_version": "2025-06-18",
      "license_status": "no redistribution approval inferred",
      "original_content_sha256": null,
      "hash_status": "original_bytes_not_archived"
    }
  ],
  "benchmark_studies": [
    {
      "study_id": "mas-study-acl-2025",
      "title": "MultiAgentBench",
      "publisher": "ACL 2025",
      "source_id": "mas-src-acl-2025",
      "task_families": [
        "interactive cooperation",
        "competition",
        "research scenario"
      ],
      "dataset_version": null,
      "metric_definitions": [
        "task score",
        "milestone achievement"
      ],
      "tested_topologies": [
        "star",
        "chain",
        "tree",
        "graph"
      ],
      "evaluation_design": "interactive LLM-agent benchmark; details require full paper for run-level reproduction",
      "limitations": "No independent AgentenCode reproduction; abstract-level observations are not deployable benchmark scores",
      "verification": "reviewed",
      "evidence_origin": "peer_reviewed_external"
    },
    {
      "study_id": "mas-study-agentbench-2024",
      "title": "AgentBench",
      "publisher": "ICLR 2024 / THUDM",
      "source_id": "mas-src-agentbench",
      "task_families": [
        "interactive single-agent and LLM agent tasks"
      ],
      "dataset_version": null,
      "metric_definitions": [],
      "tested_topologies": [],
      "evaluation_design": "broad LLM-as-agent evaluation, not orchestration protocol A/B test",
      "limitations": "Not directly comparable to MultiAgentBench on orchestration rankings",
      "verification": "reviewed",
      "evidence_origin": "independent_external"
    }
  ],
  "evaluation_results": [
    {
      "result_id": "mas-result-acl-research-graph-001",
      "study_id": "mas-study-acl-2025",
      "source_id": "mas-src-acl-2025",
      "topology": "graph",
      "task_family": "research scenario",
      "metric": "comparative task performance",
      "metric_unit": null,
      "value": null,
      "reported_observation": "The paper abstract reports graph as the best of the tested coordination protocols in its research scenario.",
      "uncertainty": null,
      "model_or_system": "multi-agent configurations described by Zhu et al.",
      "test_configuration": "abstract-level; numeric conditions not extracted",
      "verification": "reviewed",
      "evidence_origin": "peer_reviewed_external",
      "measured_by_agentencode": false,
      "direct_numeric_comparison_allowed": false,
      "source_section": "ACL Anthology abstract"
    },
    {
      "result_id": "mas-result-acl-planning-002",
      "study_id": "mas-study-acl-2025",
      "source_id": "mas-src-acl-2025",
      "topology": "cognitive planning",
      "task_family": "milestone achievement across tested scenarios",
      "metric": "milestone achievement improvement",
      "metric_unit": "as described by publisher, percentage basis not independently extracted",
      "value": null,
      "reported_observation": "Abstract states a 3% milestone achievement improvement with cognitive planning; avoid treating as verified percentage points or unrelated datasets.",
      "uncertainty": null,
      "model_or_system": "multi-agent configurations in the paper",
      "test_configuration": "abstract-level description; full run settings not re-executed",
      "verification": "reviewed",
      "evidence_origin": "peer_reviewed_external",
      "measured_by_agentencode": false,
      "direct_numeric_comparison_allowed": false,
      "source_section": "ACL Anthology abstract"
    }
  ],
  "architecture_patterns": [
    {
      "pattern_id": "mas-pattern-single",
      "slug": "single",
      "name_en": "Single-agent baseline",
      "name_de": "Einzelagent als Baseline",
      "control_flow_en": "One bounded execution path",
      "control_flow_de": "Ein begrenzter Ausführungspfad",
      "state_model_en": "No transfer required; record prompt, tools and budget",
      "state_model_de": "Keine Übergabe, Prompt/Werkzeuge/Budget dokumentieren",
      "known_risk_en": "Minimal overhead; limited specialization",
      "known_risk_de": "Wenig Overhead; begrenzte Spezialisierung",
      "verification": "pending",
      "evidence_origin": null,
      "source_id": null,
      "is_conceptual_pattern": true
    },
    {
      "pattern_id": "mas-pattern-sequential",
      "slug": "sequential",
      "name_en": "Sequential",
      "name_de": "Sequenziell",
      "control_flow_en": "Fixed ordered steps",
      "control_flow_de": "Feste Reihenfolge",
      "state_model_en": "Explicit versioned artifacts between agents",
      "state_model_de": "Versionierte Artefakte zwischen Agenten",
      "known_risk_en": "Serial latency and error propagation",
      "known_risk_de": "Serielle Latenz und Fehlerweitergabe",
      "verification": "reviewed",
      "evidence_origin": "project_documentation",
      "source_id": "mas-src-ms-orch",
      "is_conceptual_pattern": true
    },
    {
      "pattern_id": "mas-pattern-concurrent",
      "slug": "concurrent",
      "name_en": "Concurrent",
      "name_de": "Parallel",
      "control_flow_en": "Independent workers with collector",
      "control_flow_de": "Unabhängige Bearbeiter mit Sammelstelle",
      "state_model_en": "Isolated task outputs reconciled on completion",
      "state_model_de": "Getrennte Ergebnisse werden anschließend abgeglichen",
      "known_risk_en": "Merge conflicts and higher invocation cost",
      "known_risk_de": "Zusammenführungskonflikte und zusätzliche Kosten",
      "verification": "reviewed",
      "evidence_origin": "project_documentation",
      "source_id": "mas-src-ms-orch",
      "is_conceptual_pattern": true
    },
    {
      "pattern_id": "mas-pattern-handoff",
      "slug": "handoff",
      "name_en": "Handoff",
      "name_de": "Übergabe",
      "control_flow_en": "Ownership changes according to routing",
      "control_flow_de": "Zuständigkeit wechselt nach Routing",
      "state_model_en": "Task handoff includes constraints and provenance",
      "state_model_de": "Übergabe enthält Einschränkungen und Provenienz",
      "known_risk_en": "Lost context or permission ambiguity",
      "known_risk_de": "Kontextverlust oder unklare Berechtigungen",
      "verification": "reviewed",
      "evidence_origin": "project_documentation",
      "source_id": "mas-src-ms-orch",
      "is_conceptual_pattern": true
    },
    {
      "pattern_id": "mas-pattern-group_chat",
      "slug": "group_chat",
      "name_en": "Group Chat",
      "name_de": "Gruppendiskussion",
      "control_flow_en": "Agents exchange messages in shared thread",
      "control_flow_de": "Agenten tauschen Nachrichten in gemeinsamem Verlauf aus",
      "state_model_en": "Shared discussion and moderator/decision gate",
      "state_model_de": "Gemeinsamer Verlauf plus Moderation/Entscheidung",
      "known_risk_en": "Loops, token cost, false consensus",
      "known_risk_de": "Schleifen, Tokenkosten, Scheinkonsens",
      "verification": "reviewed",
      "evidence_origin": "project_documentation",
      "source_id": "mas-src-ms-orch",
      "is_conceptual_pattern": true
    },
    {
      "pattern_id": "mas-pattern-hierarchical",
      "slug": "hierarchical",
      "name_en": "Hierarchical / supervisor",
      "name_de": "Hierarchisch / Supervisor",
      "control_flow_en": "Central coordinator delegates and validates",
      "control_flow_de": "Zentrale Instanz delegiert und prüft",
      "state_model_en": "Supervisor retains assignments and approval control",
      "state_model_de": "Supervisor verantwortet Zuweisung und Freigabe",
      "known_risk_en": "Bottleneck and concentrated risk",
      "known_risk_de": "Flaschenhals und konzentriertes Risiko",
      "verification": "pending",
      "evidence_origin": null,
      "source_id": null,
      "is_conceptual_pattern": true
    },
    {
      "pattern_id": "mas-pattern-graph",
      "slug": "graph",
      "name_en": "Graph / hybrid",
      "name_de": "Graph / Hybrid",
      "control_flow_en": "Condition-based transitions and retries",
      "control_flow_de": "Bedingte Zustandsübergänge und Wiederholung",
      "state_model_en": "Versioned checkpoints and guarded state transitions",
      "state_model_de": "Versionierte Checkpoints und gesicherte Zustände",
      "known_risk_en": "Complex state debugging and loops",
      "known_risk_de": "Komplexe Zustandsanalyse und Schleifen",
      "verification": "reviewed",
      "evidence_origin": "peer_reviewed_external",
      "source_id": "mas-src-acl-2025",
      "is_conceptual_pattern": true
    }
  ],
  "evidence_links": [
    {
      "claim_id": "mas-claim-acl-protocols",
      "field": "research.coordination_protocols",
      "value": [
        "star",
        "chain",
        "tree",
        "graph"
      ],
      "source_id": "mas-src-acl-2025",
      "source_section": "Abstract",
      "scope": "MultiAgentBench interactive scenarios",
      "evidence_origin": "peer_reviewed_external",
      "verification": "reviewed",
      "checked_at": "2026-10-09"
    },
    {
      "claim_id": "mas-claim-ms-patterns",
      "field": "framework.orchestration_patterns",
      "value": [
        "sequential",
        "concurrent",
        "handoff",
        "group_chat",
        "magentic"
      ],
      "source_id": "mas-src-ms-orch",
      "source_section": "Workflow orchestrations documentation",
      "scope": "Microsoft Agent Framework workflows; not all Microsoft agents",
      "evidence_origin": "project_documentation",
      "verification": "reviewed",
      "checked_at": "2026-10-09"
    },
    {
      "claim_id": "mas-claim-a2a-version",
      "field": "protocol.specification_major_minor",
      "value": "1.0",
      "source_id": "mas-src-a2a-spec",
      "source_section": "Versioning",
      "scope": "A2A current specification as inspected on 2026-10-09",
      "evidence_origin": "project_documentation",
      "verification": "reviewed",
      "checked_at": "2026-10-09"
    },
    {
      "claim_id": "mas-claim-mcp-function",
      "field": "protocol.purpose",
      "value": "tools_resources_context_integration",
      "source_id": "mas-src-mcp-spec",
      "source_section": "Specification overview",
      "scope": "MCP specification 2025-06-18",
      "evidence_origin": "project_documentation",
      "verification": "reviewed",
      "checked_at": "2026-10-09"
    },
    {
      "claim_id": "mas-claim-marble-repo",
      "field": "study.reference_implementation",
      "value": "MARBLE",
      "source_id": "mas-src-marble",
      "source_section": "README",
      "scope": "Research code repository; not AgentenCode run",
      "evidence_origin": "project_documentation",
      "verification": "reviewed",
      "checked_at": "2026-10-09"
    },
    {
      "claim_id": "mas-claim-agentbench-scope",
      "field": "benchmark.test_target",
      "value": "LLM_agents_in_interactive_environments",
      "source_id": "mas-src-agentbench",
      "source_section": "Project README",
      "scope": "AgentBench; not a multi-agent coordination head-to-head",
      "evidence_origin": "project_documentation",
      "verification": "reviewed",
      "checked_at": "2026-10-09"
    }
  ],
  "source_watch_candidates": [
    {
      "source_id": "mas-src-acl-2025",
      "monitor_status": "candidate_not_activated",
      "review_required": true,
      "reason": "Subject to source-state, license and rate-limit review; existing AgentenWache job unchanged"
    },
    {
      "source_id": "mas-src-marble",
      "monitor_status": "candidate_not_activated",
      "review_required": true,
      "reason": "Subject to source-state, license and rate-limit review; existing AgentenWache job unchanged"
    },
    {
      "source_id": "mas-src-agentbench",
      "monitor_status": "candidate_not_activated",
      "review_required": true,
      "reason": "Subject to source-state, license and rate-limit review; existing AgentenWache job unchanged"
    },
    {
      "source_id": "mas-src-ms-orch",
      "monitor_status": "candidate_not_activated",
      "review_required": true,
      "reason": "Subject to source-state, license and rate-limit review; existing AgentenWache job unchanged"
    },
    {
      "source_id": "mas-src-a2a-spec",
      "monitor_status": "candidate_not_activated",
      "review_required": true,
      "reason": "Subject to source-state, license and rate-limit review; existing AgentenWache job unchanged"
    },
    {
      "source_id": "mas-src-mcp-spec",
      "monitor_status": "candidate_not_activated",
      "review_required": true,
      "reason": "Subject to source-state, license and rate-limit review; existing AgentenWache job unchanged"
    }
  ]
}