{
  "mode": "offline-synthetic-context-budget-diagnostics",
  "baseline_commit": "c757179c22e6104df06a44e8ce66e5217f1d96c7",
  "candidate": "working tree at invocation; published with its source commit",
  "baseline_source_sha256": {
    "src/apixly_jev_harness/__init__.py": "d3701078fca6c5e7ae6c2fffeb1f00a327dae07265b73c94aa99e0a3e16fd4d4",
    "src/apixly_jev_harness/backend.py": "e4b4c80d88c63648c1392f41eb8796430fcbce6f81d1061c71519d0759432eda",
    "src/apixly_jev_harness/callbacks.py": "2d27ad4540d910354c54de7566546b4c9d99b5a5fa87705a1905885349078497",
    "src/apixly_jev_harness/cli.py": "8d9434bf8a8b539163194a84e3f099761d08ef32201f99cbce46c139f3726d94",
    "src/apixly_jev_harness/context.py": "e0deafa0addd0f96b422a497dae8176d059dd36635954d6770c0cf61ba8c3593",
    "src/apixly_jev_harness/diagnostics.py": "24950b4d834d9e6c902996c7bcddfa208944e42300ae062e9ebc2a44f75a091f",
    "src/apixly_jev_harness/engine.py": "cdedf1ef3a769cc1557dfa63cf6b6e6c4b48a14eb7b0fc12d3d9a957a53b3be2",
    "src/apixly_jev_harness/output.py": "94be8b482e4279e29aca128a35caf54744dbe8db8188f14905488f8ddde93a0b",
    "src/apixly_jev_harness/surfaces.py": "65ce73edfe9f6a3c16de373b3a1b85671c3d2cc661f3a50dd906a15721c7d8f1",
    "src/apixly_jev_harness/task.py": "0f5801e9fdb5626e1db65d269c715e8bcf692ebaa3a91f46b9945158b4bbb6db",
    "src/apixly_jev_harness/task_cli.py": "e7ae2e5c9d6c4179997568774769fbcab292c246f8ee3bf8031c61e4f6a03f37",
    "src/apixly_jev_harness/verify.py": "97d2a178902291094f4e16db08732748d234b8c5ad8dd49a55607c8d195f1ed2"
  },
  "candidate_source_sha256": {
    "src/apixly_jev_harness/__init__.py": "10df257ec1770fce69e97a0b5199d9b034092f5557f6a6a1fb2aa61b42e7efe2",
    "src/apixly_jev_harness/backend.py": "e4b4c80d88c63648c1392f41eb8796430fcbce6f81d1061c71519d0759432eda",
    "src/apixly_jev_harness/callbacks.py": "2d27ad4540d910354c54de7566546b4c9d99b5a5fa87705a1905885349078497",
    "src/apixly_jev_harness/cli.py": "8d9434bf8a8b539163194a84e3f099761d08ef32201f99cbce46c139f3726d94",
    "src/apixly_jev_harness/context.py": "b9d76abb33f8a1ebe8731fe6d5fe6811445132a6046448fc0e2cc871507d50ea",
    "src/apixly_jev_harness/diagnostics.py": "24950b4d834d9e6c902996c7bcddfa208944e42300ae062e9ebc2a44f75a091f",
    "src/apixly_jev_harness/engine.py": "19d809ee7af67586b03a472af49908f6ec278231285d16db161a540b3f0247a6",
    "src/apixly_jev_harness/output.py": "94be8b482e4279e29aca128a35caf54744dbe8db8188f14905488f8ddde93a0b",
    "src/apixly_jev_harness/surfaces.py": "65ce73edfe9f6a3c16de373b3a1b85671c3d2cc661f3a50dd906a15721c7d8f1",
    "src/apixly_jev_harness/task.py": "0f5801e9fdb5626e1db65d269c715e8bcf692ebaa3a91f46b9945158b4bbb6db",
    "src/apixly_jev_harness/task_cli.py": "bc2aeea2af51fe26efc4528efc9f7849078030a1cea1de2420ff1e98dcabf951",
    "src/apixly_jev_harness/verify.py": "97d2a178902291094f4e16db08732748d234b8c5ad8dd49a55607c8d195f1ed2"
  },
  "runner_sha256": "ab40d8bf91dcbb06d1ae97744d2ea12cf5d935cdf7db6553e835531e7b52d6f2",
  "python_version": "3.14.2",
  "cases": {
    "ascii_projection_over": {
      "character": "A",
      "target": "context_projection",
      "characters": 15001
    },
    "ascii_projection_exact": {
      "character": "A",
      "target": "context_projection",
      "characters": 15000
    },
    "ascii_tool_over": {
      "character": "A",
      "target": "tool_selection",
      "characters": 15001
    },
    "ascii_tool_exact": {
      "character": "A",
      "target": "tool_selection",
      "characters": 15000
    },
    "ascii_tool_under": {
      "character": "A",
      "target": "tool_selection",
      "characters": 14999
    },
    "chinese_tool_exact": {
      "character": "\u4e2d",
      "target": "tool_selection",
      "characters": 15000
    },
    "emoji_tool_exact": {
      "character": "\ud83d\ude42",
      "target": "tool_selection",
      "characters": 15000
    },
    "chinese_tool_over": {
      "character": "\u4e2d",
      "target": "tool_selection",
      "characters": 15001
    },
    "emoji_projection_over": {
      "character": "\ud83d\ude42",
      "target": "context_projection",
      "characters": 15001
    },
    "ascii_parameter_exact": {
      "character": "A",
      "target": "parameter_selection",
      "characters": 15000
    },
    "ascii_parameter_over": {
      "character": "A",
      "target": "parameter_selection",
      "characters": 15001
    },
    "chinese_parameter_exact": {
      "character": "\u4e2d",
      "target": "parameter_selection",
      "characters": 15000
    },
    "direct_ascii_over": {
      "character": "A",
      "target": "direct_decide",
      "characters": 15001
    },
    "direct_chinese_over": {
      "character": "\u4e2d",
      "target": "direct_decide",
      "characters": 15001
    }
  },
  "behavior_parity": true,
  "reproduction": "python benchmarks/context_budget.py --output local-results/context-budget-ab.json",
  "scope": "Fourteen fixed ASCII/Chinese/emoji shared-context fixtures, three paired repeats with alternating arm order. Custom policy, exact scalar payload sizes and a loop-only frozen clock make projection/tool/parameter thresholds deterministic. The 15,000 limit remains JSON characters from json.dumps(ensure_ascii=False) with default separators and rejects only greater values. Unicode contexts whose UTF-8 size exceeds 15,000 still admit when character count fits. Parameter excess occurs after a scripted tool choice but before intent/effect. Direct Engine.decide remains outside this loop limit. All decisions are scripted; native inference entrypoints are forbidden, actual model/UI/network calls zero unless the guard is hit.",
  "negative_results": "Fixed failure diagnostics and successful call counters add returned context and local measurement work. This is not a UTF-8 byte, token, candidate text, question, full provider prompt or direct Engine.decide budget. Direct decide still has upstream analysis-spec validation: the first synthetic smoke with 16,000 shared characters failed its 16,000-character full-spec cap after default questions were added; direct fixtures therefore use 15,001 shared characters to isolate the loop boundary. A custom policy omits ordinary production history and the frozen clock excludes wall-budget races. No truncation, context-quality improvement, model accuracy, speed or fee benefit is claimed. Static scalar context measurements are diagnostic evidence and do not authorize execution.",
  "cost_conversion": "Zero actual model usage only with no native guard attempts; a guard hit records unknown usage and fails the fixture. No tariff or savings assumed. Real whole-operation timing uses perf_counter and includes synthetic registration, payload/projection setup, loop admission, scripted decisions, tool effects and private journal serialization; excludes process/import setup and terminal-analysis/source-hashing. Frozen monotonic values stabilize only the synthetic remaining-time context and do not stand in for end-to-end timing. Returned context bytes exclude the local archive path and never include raw payloads; request metrics are shared-context scalar sizes, not provider prompt/token counts.",
  "summary": {
    "baseline": {
      "runs": 42,
      "correct": 6,
      "behavior_correct": 42,
      "failure_metadata_correct": 24,
      "success_metadata_correct": 21,
      "statuses": {
        "admitted": 6,
        "done": 18,
        "needs_review": 18
      },
      "scripted_decisions": 33,
      "effects": 18,
      "intent_events": 18,
      "native_inference_attempts": 0,
      "terminal_marker_exposures": 0,
      "median_whole_operation_seconds": 0.001507729000877589,
      "returned_context_bytes": 15414,
      "request_context_characters": 492081,
      "request_context_utf8_bytes": 979227,
      "actual_model_usage": {
        "requests": 0,
        "models": {},
        "input_tokens": 0,
        "output_tokens": 0,
        "usage_complete": true
      }
    },
    "candidate": {
      "runs": 42,
      "correct": 42,
      "behavior_correct": 42,
      "failure_metadata_correct": 42,
      "success_metadata_correct": 42,
      "statuses": {
        "admitted": 6,
        "done": 18,
        "needs_review": 18
      },
      "scripted_decisions": 33,
      "effects": 18,
      "intent_events": 18,
      "native_inference_attempts": 0,
      "terminal_marker_exposures": 0,
      "median_whole_operation_seconds": 0.0015135209978325292,
      "returned_context_bytes": 19692,
      "request_context_characters": 492081,
      "request_context_utf8_bytes": 979227,
      "actual_model_usage": {
        "requests": 0,
        "models": {},
        "input_tokens": 0,
        "output_tokens": 0,
        "usage_complete": true
      }
    }
  },
  "rows": [
    {
      "case": "ascii_projection_over",
      "runtime_version": "0.1.8",
      "runtime_loaded_from_requested_source": true,
      "fixture": {
        "character": "A",
        "target": "context_projection",
        "characters": 15001,
        "marker_prefix_characters": 24,
        "payload_character_repetitions": 14899,
        "frozen_monotonic": 100.0,
        "max_steps": 2,
        "timeout_seconds": 120
      },
      "projected_context_sizes": {
        "context_projection": {
          "characters": 15001,
          "utf8_bytes": 15001
        },
        "tool_selection": {
          "characters": 15037,
          "utf8_bytes": 15037
        }
      },
      "expected": {
        "status": "needs_review",
        "scripted_decisions": 0,
        "effects": 0,
        "intent_events": 0,
        "reason": "context_budget_exceeded",
        "context_budget": {
          "scope": "loop_shared_context",
          "stage": "context_projection",
          "unit": "json_characters",
          "limit": 15000,
          "observed": 15001
        }
      },
      "result": {
        "status": "needs_review",
        "reason": "context_budget_exceeded",
        "context_budget": null,
        "failure_metadata_correct": false,
        "success_metadata_correct": true,
        "behavior_correct": true,
        "scripted_decisions": 0,
        "effects": 0,
        "intent_events": 0,
        "native_inference_attempts": 0,
        "request_context_sizes": [],
        "returned_call_telemetry": [],
        "terminal_marker_exposures": 0,
        "reported_scripted_usage_complete": true
      },
      "correct": false,
      "whole_operation_seconds": 0.0010069160052808002,
      "returned_context_bytes": 259,
      "actual_model_usage": {
        "requests": 0,
        "models": {},
        "input_tokens": 0,
        "output_tokens": 0,
        "usage_complete": true
      },
      "arm": "baseline",
      "repeat": 1
    },
    {
      "case": "ascii_projection_over",
      "runtime_version": "0.1.9",
      "runtime_loaded_from_requested_source": true,
      "fixture": {
        "character": "A",
        "target": "context_projection",
        "characters": 15001,
        "marker_prefix_characters": 24,
        "payload_character_repetitions": 14899,
        "frozen_monotonic": 100.0,
        "max_steps": 2,
        "timeout_seconds": 120
      },
      "projected_context_sizes": {
        "context_projection": {
          "characters": 15001,
          "utf8_bytes": 15001
        },
        "tool_selection": {
          "characters": 15037,
          "utf8_bytes": 15037
        }
      },
      "expected": {
        "status": "needs_review",
        "scripted_decisions": 0,
        "effects": 0,
        "intent_events": 0,
        "reason": "context_budget_exceeded",
        "context_budget": {
          "scope": "loop_shared_context",
          "stage": "context_projection",
          "unit": "json_characters",
          "limit": 15000,
          "observed": 15001
        }
      },
      "result": {
        "status": "needs_review",
        "reason": "context_budget_exceeded",
        "context_budget": {
          "scope": "loop_shared_context",
          "stage": "context_projection",
          "unit": "json_characters",
          "limit": 15000,
          "observed": 15001
        },
        "failure_metadata_correct": true,
        "success_metadata_correct": true,
        "behavior_correct": true,
        "scripted_decisions": 0,
        "effects": 0,
        "intent_events": 0,
        "native_inference_attempts": 0,
        "request_context_sizes": [],
        "returned_call_telemetry": [],
        "terminal_marker_exposures": 0,
        "reported_scripted_usage_complete": true
      },
      "correct": true,
      "whole_operation_seconds": 0.001057458997820504,
      "returned_context_bytes": 404,
      "actual_model_usage": {
        "requests": 0,
        "models": {},
        "input_tokens": 0,
        "output_tokens": 0,
        "usage_complete": true
      },
      "arm": "candidate",
      "repeat": 1
    },
    {
      "case": "ascii_projection_over",
      "runtime_version": "0.1.9",
      "runtime_loaded_from_requested_source": true,
      "fixture": {
        "character": "A",
        "target": "context_projection",
        "characters": 15001,
        "marker_prefix_characters": 24,
        "payload_character_repetitions": 14899,
        "frozen_monotonic": 100.0,
        "max_steps": 2,
        "timeout_seconds": 120
      },
      "projected_context_sizes": {
        "context_projection": {
          "characters": 15001,
          "utf8_bytes": 15001
        },
        "tool_selection": {
          "characters": 15037,
          "utf8_bytes": 15037
        }
      },
      "expected": {
        "status": "needs_review",
        "scripted_decisions": 0,
        "effects": 0,
        "intent_events": 0,
        "reason": "context_budget_exceeded",
        "context_budget": {
          "scope": "loop_shared_context",
          "stage": "context_projection",
          "unit": "json_characters",
          "limit": 15000,
          "observed": 15001
        }
      },
      "result": {
        "status": "needs_review",
        "reason": "context_budget_exceeded",
        "context_budget": {
          "scope": "loop_shared_context",
          "stage": "context_projection",
          "unit": "json_characters",
          "limit": 15000,
          "observed": 15001
        },
        "failure_metadata_correct": true,
        "success_metadata_correct": true,
        "behavior_correct": true,
        "scripted_decisions": 0,
        "effects": 0,
        "intent_events": 0,
        "native_inference_attempts": 0,
        "request_context_sizes": [],
        "returned_call_telemetry": [],
        "terminal_marker_exposures": 0,
        "reported_scripted_usage_complete": true
      },
      "correct": true,
      "whole_operation_seconds": 0.0010738330020103604,
      "returned_context_bytes": 404,
      "actual_model_usage": {
        "requests": 0,
        "models": {},
        "input_tokens": 0,
        "output_tokens": 0,
        "usage_complete": true
      },
      "arm": "candidate",
      "repeat": 2
    },
    {
      "case": "ascii_projection_over",
      "runtime_version": "0.1.8",
      "runtime_loaded_from_requested_source": true,
      "fixture": {
        "character": "A",
        "target": "context_projection",
        "characters": 15001,
        "marker_prefix_characters": 24,
        "payload_character_repetitions": 14899,
        "frozen_monotonic": 100.0,
        "max_steps": 2,
        "timeout_seconds": 120
      },
      "projected_context_sizes": {
        "context_projection": {
          "characters": 15001,
          "utf8_bytes": 15001
        },
        "tool_selection": {
          "characters": 15037,
          "utf8_bytes": 15037
        }
      },
      "expected": {
        "status": "needs_review",
        "scripted_decisions": 0,
        "effects": 0,
        "intent_events": 0,
        "reason": "context_budget_exceeded",
        "context_budget": {
          "scope": "loop_shared_context",
          "stage": "context_projection",
          "unit": "json_characters",
          "limit": 15000,
          "observed": 15001
        }
      },
      "result": {
        "status": "needs_review",
        "reason": "context_budget_exceeded",
        "context_budget": null,
        "failure_metadata_correct": false,
        "success_metadata_correct": true,
        "behavior_correct": true,
        "scripted_decisions": 0,
        "effects": 0,
        "intent_events": 0,
        "native_inference_attempts": 0,
        "request_context_sizes": [],
        "returned_call_telemetry": [],
        "terminal_marker_exposures": 0,
        "reported_scripted_usage_complete": true
      },
      "correct": false,
      "whole_operation_seconds": 0.0012641250068554655,
      "returned_context_bytes": 259,
      "actual_model_usage": {
        "requests": 0,
        "models": {},
        "input_tokens": 0,
        "output_tokens": 0,
        "usage_complete": true
      },
      "arm": "baseline",
      "repeat": 2
    },
    {
      "case": "ascii_projection_over",
      "runtime_version": "0.1.8",
      "runtime_loaded_from_requested_source": true,
      "fixture": {
        "character": "A",
        "target": "context_projection",
        "characters": 15001,
        "marker_prefix_characters": 24,
        "payload_character_repetitions": 14899,
        "frozen_monotonic": 100.0,
        "max_steps": 2,
        "timeout_seconds": 120
      },
      "projected_context_sizes": {
        "context_projection": {
          "characters": 15001,
          "utf8_bytes": 15001
        },
        "tool_selection": {
          "characters": 15037,
          "utf8_bytes": 15037
        }
      },
      "expected": {
        "status": "needs_review",
        "scripted_decisions": 0,
        "effects": 0,
        "intent_events": 0,
        "reason": "context_budget_exceeded",
        "context_budget": {
          "scope": "loop_shared_context",
          "stage": "context_projection",
          "unit": "json_characters",
          "limit": 15000,
          "observed": 15001
        }
      },
      "result": {
        "status": "needs_review",
        "reason": "context_budget_exceeded",
        "context_budget": null,
        "failure_metadata_correct": false,
        "success_metadata_correct": true,
        "behavior_correct": true,
        "scripted_decisions": 0,
        "effects": 0,
        "intent_events": 0,
        "native_inference_attempts": 0,
        "request_context_sizes": [],
        "returned_call_telemetry": [],
        "terminal_marker_exposures": 0,
        "reported_scripted_usage_complete": true
      },
      "correct": false,
      "whole_operation_seconds": 0.001280625001527369,
      "returned_context_bytes": 259,
      "actual_model_usage": {
        "requests": 0,
        "models": {},
        "input_tokens": 0,
        "output_tokens": 0,
        "usage_complete": true
      },
      "arm": "baseline",
      "repeat": 3
    },
    {
      "case": "ascii_projection_over",
      "runtime_version": "0.1.9",
      "runtime_loaded_from_requested_source": true,
      "fixture": {
        "character": "A",
        "target": "context_projection",
        "characters": 15001,
        "marker_prefix_characters": 24,
        "payload_character_repetitions": 14899,
        "frozen_monotonic": 100.0,
        "max_steps": 2,
        "timeout_seconds": 120
      },
      "projected_context_sizes": {
        "context_projection": {
          "characters": 15001,
          "utf8_bytes": 15001
        },
        "tool_selection": {
          "characters": 15037,
          "utf8_bytes": 15037
        }
      },
      "expected": {
        "status": "needs_review",
        "scripted_decisions": 0,
        "effects": 0,
        "intent_events": 0,
        "reason": "context_budget_exceeded",
        "context_budget": {
          "scope": "loop_shared_context",
          "stage": "context_projection",
          "unit": "json_characters",
          "limit": 15000,
          "observed": 15001
        }
      },
      "result": {
        "status": "needs_review",
        "reason": "context_budget_exceeded",
        "context_budget": {
          "scope": "loop_shared_context",
          "stage": "context_projection",
          "unit": "json_characters",
          "limit": 15000,
          "observed": 15001
        },
        "failure_metadata_correct": true,
        "success_metadata_correct": true,
        "behavior_correct": true,
        "scripted_decisions": 0,
        "effects": 0,
        "intent_events": 0,
        "native_inference_attempts": 0,
        "request_context_sizes": [],
        "returned_call_telemetry": [],
        "terminal_marker_exposures": 0,
        "reported_scripted_usage_complete": true
      },
      "correct": true,
      "whole_operation_seconds": 0.001098791995900683,
      "returned_context_bytes": 404,
      "actual_model_usage": {
        "requests": 0,
        "models": {},
        "input_tokens": 0,
        "output_tokens": 0,
        "usage_complete": true
      },
      "arm": "candidate",
      "repeat": 3
    },
    {
      "case": "ascii_projection_exact",
      "runtime_version": "0.1.8",
      "runtime_loaded_from_requested_source": true,
      "fixture": {
        "character": "A",
        "target": "context_projection",
        "characters": 15000,
        "marker_prefix_characters": 24,
        "payload_character_repetitions": 14898,
        "frozen_monotonic": 100.0,
        "max_steps": 2,
        "timeout_seconds": 120
      },
      "projected_context_sizes": {
        "context_projection": {
          "characters": 15000,
          "utf8_bytes": 15000
        },
        "tool_selection": {
          "characters": 15036,
          "utf8_bytes": 15036
        }
      },
      "expected": {
        "status": "needs_review",
        "scripted_decisions": 0,
        "effects": 0,
        "intent_events": 0,
        "reason": "context_budget_exceeded",
        "context_budget": {
          "scope": "loop_shared_context",
          "stage": "tool_selection",
          "unit": "json_characters",
          "limit": 15000,
          "observed": 15036
        }
      },
      "result": {
        "status": "needs_review",
        "reason": "context_budget_exceeded",
        "context_budget": null,
        "failure_metadata_correct": false,
        "success_metadata_correct": true,
        "behavior_correct": true,
        "scripted_decisions": 0,
        "effects": 0,
        "intent_events": 0,
        "native_inference_attempts": 0,
        "request_context_sizes": [],
        "returned_call_telemetry": [],
        "terminal_marker_exposures": 0,
        "reported_scripted_usage_complete": true
      },
      "correct": false,
      "whole_operation_seconds": 0.0010841670009540394,
      "returned_context_bytes": 259,
      "actual_model_usage": {
        "requests": 0,
        "models": {},
        "input_tokens": 0,
        "output_tokens": 0,
        "usage_complete": true
      },
      "arm": "baseline",
      "repeat": 1
    },
    {
      "case": "ascii_projection_exact",
      "runtime_version": "0.1.9",
      "runtime_loaded_from_requested_source": true,
      "fixture": {
        "character": "A",
        "target": "context_projection",
        "characters": 15000,
        "marker_prefix_characters": 24,
        "payload_character_repetitions": 14898,
        "frozen_monotonic": 100.0,
        "max_steps": 2,
        "timeout_seconds": 120
      },
      "projected_context_sizes": {
        "context_projection": {
          "characters": 15000,
          "utf8_bytes": 15000
        },
        "tool_selection": {
          "characters": 15036,
          "utf8_bytes": 15036
        }
      },
      "expected": {
        "status": "needs_review",
        "scripted_decisions": 0,
        "effects": 0,
        "intent_events": 0,
        "reason": "context_budget_exceeded",
        "context_budget": {
          "scope": "loop_shared_context",
          "stage": "tool_selection",
          "unit": "json_characters",
          "limit": 15000,
          "observed": 15036
        }
      },
      "result": {
        "status": "needs_review",
        "reason": "context_budget_exceeded",
        "context_budget": {
          "scope": "loop_shared_context",
          "stage": "tool_selection",
          "unit": "json_characters",
          "limit": 15000,
          "observed": 15036
        },
        "failure_metadata_correct": true,
        "success_metadata_correct": true,
        "behavior_correct": true,
        "scripted_decisions": 0,
        "effects": 0,
        "intent_events": 0,
        "native_inference_attempts": 0,
        "request_context_sizes": [],
        "returned_call_telemetry": [],
        "terminal_marker_exposures": 0,
        "reported_scripted_usage_complete": true
      },
      "correct": true,
      "whole_operation_seconds": 0.0011110420018667355,
      "returned_context_bytes": 400,
      "actual_model_usage": {
        "requests": 0,
        "models": {},
        "input_tokens": 0,
        "output_tokens": 0,
        "usage_complete": true
      },
      "arm": "candidate",
      "repeat": 1
    },
    {
      "case": "ascii_projection_exact",
      "runtime_version": "0.1.9",
      "runtime_loaded_from_requested_source": true,
      "fixture": {
        "character": "A",
        "target": "context_projection",
        "characters": 15000,
        "marker_prefix_characters": 24,
        "payload_character_repetitions": 14898,
        "frozen_monotonic": 100.0,
        "max_steps": 2,
        "timeout_seconds": 120
      },
      "projected_context_sizes": {
        "context_projection": {
          "characters": 15000,
          "utf8_bytes": 15000
        },
        "tool_selection": {
          "characters": 15036,
          "utf8_bytes": 15036
        }
      },
      "expected": {
        "status": "needs_review",
        "scripted_decisions": 0,
        "effects": 0,
        "intent_events": 0,
        "reason": "context_budget_exceeded",
        "context_budget": {
          "scope": "loop_shared_context",
          "stage": "tool_selection",
          "unit": "json_characters",
          "limit": 15000,
          "observed": 15036
        }
      },
      "result": {
        "status": "needs_review",
        "reason": "context_budget_exceeded",
        "context_budget": {
          "scope": "loop_shared_context",
          "stage": "tool_selection",
          "unit": "json_characters",
          "limit": 15000,
          "observed": 15036
        },
        "failure_metadata_correct": true,
        "success_metadata_correct": true,
        "behavior_correct": true,
        "scripted_decisions": 0,
        "effects": 0,
        "intent_events": 0,
        "native_inference_attempts": 0,
        "request_context_sizes": [],
        "returned_call_telemetry": [],
        "terminal_marker_exposures": 0,
        "reported_scripted_usage_complete": true
      },
      "correct": true,
      "whole_operation_seconds": 0.0010995410120813176,
      "returned_context_bytes": 400,
      "actual_model_usage": {
        "requests": 0,
        "models": {},
        "input_tokens": 0,
        "output_tokens": 0,
        "usage_complete": true
      },
      "arm": "candidate",
      "repeat": 2
    },
    {
      "case": "ascii_projection_exact",
      "runtime_version": "0.1.8",
      "runtime_loaded_from_requested_source": true,
      "fixture": {
        "character": "A",
        "target": "context_projection",
        "characters": 15000,
        "marker_prefix_characters": 24,
        "payload_character_repetitions": 14898,
        "frozen_monotonic": 100.0,
        "max_steps": 2,
        "timeout_seconds": 120
      },
      "projected_context_sizes": {
        "context_projection": {
          "characters": 15000,
          "utf8_bytes": 15000
        },
        "tool_selection": {
          "characters": 15036,
          "utf8_bytes": 15036
        }
      },
      "expected": {
        "status": "needs_review",
        "scripted_decisions": 0,
        "effects": 0,
        "intent_events": 0,
        "reason": "context_budget_exceeded",
        "context_budget": {
          "scope": "loop_shared_context",
          "stage": "tool_selection",
          "unit": "json_characters",
          "limit": 15000,
          "observed": 15036
        }
      },
      "result": {
        "status": "needs_review",
        "reason": "context_budget_exceeded",
        "context_budget": null,
        "failure_metadata_correct": false,
        "success_metadata_correct": true,
        "behavior_correct": true,
        "scripted_decisions": 0,
        "effects": 0,
        "intent_events": 0,
        "native_inference_attempts": 0,
        "request_context_sizes": [],
        "returned_call_telemetry": [],
        "terminal_marker_exposures": 0,
        "reported_scripted_usage_complete": true
      },
      "correct": false,
      "whole_operation_seconds": 0.001435666999896057,
      "returned_context_bytes": 259,
      "actual_model_usage": {
        "requests": 0,
        "models": {},
        "input_tokens": 0,
        "output_tokens": 0,
        "usage_complete": true
      },
      "arm": "baseline",
      "repeat": 2
    },
    {
      "case": "ascii_projection_exact",
      "runtime_version": "0.1.8",
      "runtime_loaded_from_requested_source": true,
      "fixture": {
        "character": "A",
        "target": "context_projection",
        "characters": 15000,
        "marker_prefix_characters": 24,
        "payload_character_repetitions": 14898,
        "frozen_monotonic": 100.0,
        "max_steps": 2,
        "timeout_seconds": 120
      },
      "projected_context_sizes": {
        "context_projection": {
          "characters": 15000,
          "utf8_bytes": 15000
        },
        "tool_selection": {
          "characters": 15036,
          "utf8_bytes": 15036
        }
      },
      "expected": {
        "status": "needs_review",
        "scripted_decisions": 0,
        "effects": 0,
        "intent_events": 0,
        "reason": "context_budget_exceeded",
        "context_budget": {
          "scope": "loop_shared_context",
          "stage": "tool_selection",
          "unit": "json_characters",
          "limit": 15000,
          "observed": 15036
        }
      },
      "result": {
        "status": "needs_review",
        "reason": "context_budget_exceeded",
        "context_budget": null,
        "failure_metadata_correct": false,
        "success_metadata_correct": true,
        "behavior_correct": true,
        "scripted_decisions": 0,
        "effects": 0,
        "intent_events": 0,
        "native_inference_attempts": 0,
        "request_context_sizes": [],
        "returned_call_telemetry": [],
        "terminal_marker_exposures": 0,
        "reported_scripted_usage_complete": true
      },
      "correct": false,
      "whole_operation_seconds": 0.0015645419916836545,
      "returned_context_bytes": 259,
      "actual_model_usage": {
        "requests": 0,
        "models": {},
        "input_tokens": 0,
        "output_tokens": 0,
        "usage_complete": true
      },
      "arm": "baseline",
      "repeat": 3
    },
    {
      "case": "ascii_projection_exact",
      "runtime_version": "0.1.9",
      "runtime_loaded_from_requested_source": true,
      "fixture": {
        "character": "A",
        "target": "context_projection",
        "characters": 15000,
        "marker_prefix_characters": 24,
        "payload_character_repetitions": 14898,
        "frozen_monotonic": 100.0,
        "max_steps": 2,
        "timeout_seconds": 120
      },
      "projected_context_sizes": {
        "context_projection": {
          "characters": 15000,
          "utf8_bytes": 15000
        },
        "tool_selection": {
          "characters": 15036,
          "utf8_bytes": 15036
        }
      },
      "expected": {
        "status": "needs_review",
        "scripted_decisions": 0,
        "effects": 0,
        "intent_events": 0,
        "reason": "context_budget_exceeded",
        "context_budget": {
          "scope": "loop_shared_context",
          "stage": "tool_selection",
          "unit": "json_characters",
          "limit": 15000,
          "observed": 15036
        }
      },
      "result": {
        "status": "needs_review",
        "reason": "context_budget_exceeded",
        "context_budget": {
          "scope": "loop_shared_context",
          "stage": "tool_selection",
          "unit": "json_characters",
          "limit": 15000,
          "observed": 15036
        },
        "failure_metadata_correct": true,
        "success_metadata_correct": true,
        "behavior_correct": true,
        "scripted_decisions": 0,
        "effects": 0,
        "intent_events": 0,
        "native_inference_attempts": 0,
        "request_context_sizes": [],
        "returned_call_telemetry": [],
        "terminal_marker_exposures": 0,
        "reported_scripted_usage_complete": true
      },
      "correct": true,
      "whole_operation_seconds": 0.0013225829898146912,
      "returned_context_bytes": 400,
      "actual_model_usage": {
        "requests": 0,
        "models": {},
        "input_tokens": 0,
        "output_tokens": 0,
        "usage_complete": true
      },
      "arm": "candidate",
      "repeat": 3
    },
    {
      "case": "ascii_tool_over",
      "runtime_version": "0.1.8",
      "runtime_loaded_from_requested_source": true,
      "fixture": {
        "character": "A",
        "target": "tool_selection",
        "characters": 15001,
        "marker_prefix_characters": 24,
        "payload_character_repetitions": 14863,
        "frozen_monotonic": 100.0,
        "max_steps": 2,
        "timeout_seconds": 120
      },
      "projected_context_sizes": {
        "context_projection": {
          "characters": 14965,
          "utf8_bytes": 14965
        },
        "tool_selection": {
          "characters": 15001,
          "utf8_bytes": 15001
        }
      },
      "expected": {
        "status": "needs_review",
        "scripted_decisions": 0,
        "effects": 0,
        "intent_events": 0,
        "reason": "context_budget_exceeded",
        "context_budget": {
          "scope": "loop_shared_context",
          "stage": "tool_selection",
          "unit": "json_characters",
          "limit": 15000,
          "observed": 15001
        }
      },
      "result": {
        "status": "needs_review",
        "reason": "context_budget_exceeded",
        "context_budget": null,
        "failure_metadata_correct": false,
        "success_metadata_correct": true,
        "behavior_correct": true,
        "scripted_decisions": 0,
        "effects": 0,
        "intent_events": 0,
        "native_inference_attempts": 0,
        "request_context_sizes": [],
        "returned_call_telemetry": [],
        "terminal_marker_exposures": 0,
        "reported_scripted_usage_complete": true
      },
      "correct": false,
      "whole_operation_seconds": 0.0012527080107247457,
      "returned_context_bytes": 259,
      "actual_model_usage": {
        "requests": 0,
        "models": {},
        "input_tokens": 0,
        "output_tokens": 0,
        "usage_complete": true
      },
      "arm": "baseline",
      "repeat": 1
    },
    {
      "case": "ascii_tool_over",
      "runtime_version": "0.1.9",
      "runtime_loaded_from_requested_source": true,
      "fixture": {
        "character": "A",
        "target": "tool_selection",
        "characters": 15001,
        "marker_prefix_characters": 24,
        "payload_character_repetitions": 14863,
        "frozen_monotonic": 100.0,
        "max_steps": 2,
        "timeout_seconds": 120
      },
      "projected_context_sizes": {
        "context_projection": {
          "characters": 14965,
          "utf8_bytes": 14965
        },
        "tool_selection": {
          "characters": 15001,
          "utf8_bytes": 15001
        }
      },
      "expected": {
        "status": "needs_review",
        "scripted_decisions": 0,
        "effects": 0,
        "intent_events": 0,
        "reason": "context_budget_exceeded",
        "context_budget": {
          "scope": "loop_shared_context",
          "stage": "tool_selection",
          "unit": "json_characters",
          "limit": 15000,
          "observed": 15001
        }
      },
      "result": {
        "status": "needs_review",
        "reason": "context_budget_exceeded",
        "context_budget": {
          "scope": "loop_shared_context",
          "stage": "tool_selection",
          "unit": "json_characters",
          "limit": 15000,
          "observed": 15001
        },
        "failure_metadata_correct": true,
        "success_metadata_correct": true,
        "behavior_correct": true,
        "scripted_decisions": 0,
        "effects": 0,
        "intent_events": 0,
        "native_inference_attempts": 0,
        "request_context_sizes": [],
        "returned_call_telemetry": [],
        "terminal_marker_exposures": 0,
        "reported_scripted_usage_complete": true
      },
      "correct": true,
      "whole_operation_seconds": 0.0012291669991100207,
      "returned_context_bytes": 400,
      "actual_model_usage": {
        "requests": 0,
        "models": {},
        "input_tokens": 0,
        "output_tokens": 0,
        "usage_complete": true
      },
      "arm": "candidate",
      "repeat": 1
    },
    {
      "case": "ascii_tool_over",
      "runtime_version": "0.1.9",
      "runtime_loaded_from_requested_source": true,
      "fixture": {
        "character": "A",
        "target": "tool_selection",
        "characters": 15001,
        "marker_prefix_characters": 24,
        "payload_character_repetitions": 14863,
        "frozen_monotonic": 100.0,
        "max_steps": 2,
        "timeout_seconds": 120
      },
      "projected_context_sizes": {
        "context_projection": {
          "characters": 14965,
          "utf8_bytes": 14965
        },
        "tool_selection": {
          "characters": 15001,
          "utf8_bytes": 15001
        }
      },
      "expected": {
        "status": "needs_review",
        "scripted_decisions": 0,
        "effects": 0,
        "intent_events": 0,
        "reason": "context_budget_exceeded",
        "context_budget": {
          "scope": "loop_shared_context",
          "stage": "tool_selection",
          "unit": "json_characters",
          "limit": 15000,
          "observed": 15001
        }
      },
      "result": {
        "status": "needs_review",
        "reason": "context_budget_exceeded",
        "context_budget": {
          "scope": "loop_shared_context",
          "stage": "tool_selection",
          "unit": "json_characters",
          "limit": 15000,
          "observed": 15001
        },
        "failure_metadata_correct": true,
        "success_metadata_correct": true,
        "behavior_correct": true,
        "scripted_decisions": 0,
        "effects": 0,
        "intent_events": 0,
        "native_inference_attempts": 0,
        "request_context_sizes": [],
        "returned_call_telemetry": [],
        "terminal_marker_exposures": 0,
        "reported_scripted_usage_complete": true
      },
      "correct": true,
      "whole_operation_seconds": 0.0012999580067116767,
      "returned_context_bytes": 400,
      "actual_model_usage": {
        "requests": 0,
        "models": {},
        "input_tokens": 0,
        "output_tokens": 0,
        "usage_complete": true
      },
      "arm": "candidate",
      "repeat": 2
    },
    {
      "case": "ascii_tool_over",
      "runtime_version": "0.1.8",
      "runtime_loaded_from_requested_source": true,
      "fixture": {
        "character": "A",
        "target": "tool_selection",
        "characters": 15001,
        "marker_prefix_characters": 24,
        "payload_character_repetitions": 14863,
        "frozen_monotonic": 100.0,
        "max_steps": 2,
        "timeout_seconds": 120
      },
      "projected_context_sizes": {
        "context_projection": {
          "characters": 14965,
          "utf8_bytes": 14965
        },
        "tool_selection": {
          "characters": 15001,
          "utf8_bytes": 15001
        }
      },
      "expected": {
        "status": "needs_review",
        "scripted_decisions": 0,
        "effects": 0,
        "intent_events": 0,
        "reason": "context_budget_exceeded",
        "context_budget": {
          "scope": "loop_shared_context",
          "stage": "tool_selection",
          "unit": "json_characters",
          "limit": 15000,
          "observed": 15001
        }
      },
      "result": {
        "status": "needs_review",
        "reason": "context_budget_exceeded",
        "context_budget": null,
        "failure_metadata_correct": false,
        "success_metadata_correct": true,
        "behavior_correct": true,
        "scripted_decisions": 0,
        "effects": 0,
        "intent_events": 0,
        "native_inference_attempts": 0,
        "request_context_sizes": [],
        "returned_call_telemetry": [],
        "terminal_marker_exposures": 0,
        "reported_scripted_usage_complete": true
      },
      "correct": false,
      "whole_operation_seconds": 0.0012080000014975667,
      "returned_context_bytes": 259,
      "actual_model_usage": {
        "requests": 0,
        "models": {},
        "input_tokens": 0,
        "output_tokens": 0,
        "usage_complete": true
      },
      "arm": "baseline",
      "repeat": 2
    },
    {
      "case": "ascii_tool_over",
      "runtime_version": "0.1.8",
      "runtime_loaded_from_requested_source": true,
      "fixture": {
        "character": "A",
        "target": "tool_selection",
        "characters": 15001,
        "marker_prefix_characters": 24,
        "payload_character_repetitions": 14863,
        "frozen_monotonic": 100.0,
        "max_steps": 2,
        "timeout_seconds": 120
      },
      "projected_context_sizes": {
        "context_projection": {
          "characters": 14965,
          "utf8_bytes": 14965
        },
        "tool_selection": {
          "characters": 15001,
          "utf8_bytes": 15001
        }
      },
      "expected": {
        "status": "needs_review",
        "scripted_decisions": 0,
        "effects": 0,
        "intent_events": 0,
        "reason": "context_budget_exceeded",
        "context_budget": {
          "scope": "loop_shared_context",
          "stage": "tool_selection",
          "unit": "json_characters",
          "limit": 15000,
          "observed": 15001
        }
      },
      "result": {
        "status": "needs_review",
        "reason": "context_budget_exceeded",
        "context_budget": null,
        "failure_metadata_correct": false,
        "success_metadata_correct": true,
        "behavior_correct": true,
        "scripted_decisions": 0,
        "effects": 0,
        "intent_events": 0,
        "native_inference_attempts": 0,
        "request_context_sizes": [],
        "returned_call_telemetry": [],
        "terminal_marker_exposures": 0,
        "reported_scripted_usage_complete": true
      },
      "correct": false,
      "whole_operation_seconds": 0.0013115829933667555,
      "returned_context_bytes": 259,
      "actual_model_usage": {
        "requests": 0,
        "models": {},
        "input_tokens": 0,
        "output_tokens": 0,
        "usage_complete": true
      },
      "arm": "baseline",
      "repeat": 3
    },
    {
      "case": "ascii_tool_over",
      "runtime_version": "0.1.9",
      "runtime_loaded_from_requested_source": true,
      "fixture": {
        "character": "A",
        "target": "tool_selection",
        "characters": 15001,
        "marker_prefix_characters": 24,
        "payload_character_repetitions": 14863,
        "frozen_monotonic": 100.0,
        "max_steps": 2,
        "timeout_seconds": 120
      },
      "projected_context_sizes": {
        "context_projection": {
          "characters": 14965,
          "utf8_bytes": 14965
        },
        "tool_selection": {
          "characters": 15001,
          "utf8_bytes": 15001
        }
      },
      "expected": {
        "status": "needs_review",
        "scripted_decisions": 0,
        "effects": 0,
        "intent_events": 0,
        "reason": "context_budget_exceeded",
        "context_budget": {
          "scope": "loop_shared_context",
          "stage": "tool_selection",
          "unit": "json_characters",
          "limit": 15000,
          "observed": 15001
        }
      },
      "result": {
        "status": "needs_review",
        "reason": "context_budget_exceeded",
        "context_budget": {
          "scope": "loop_shared_context",
          "stage": "tool_selection",
          "unit": "json_characters",
          "limit": 15000,
          "observed": 15001
        },
        "failure_metadata_correct": true,
        "success_metadata_correct": true,
        "behavior_correct": true,
        "scripted_decisions": 0,
        "effects": 0,
        "intent_events": 0,
        "native_inference_attempts": 0,
        "request_context_sizes": [],
        "returned_call_telemetry": [],
        "terminal_marker_exposures": 0,
        "reported_scripted_usage_complete": true
      },
      "correct": true,
      "whole_operation_seconds": 0.0013736669934587553,
      "returned_context_bytes": 400,
      "actual_model_usage": {
        "requests": 0,
        "models": {},
        "input_tokens": 0,
        "output_tokens": 0,
        "usage_complete": true
      },
      "arm": "candidate",
      "repeat": 3
    },
    {
      "case": "ascii_tool_exact",
      "runtime_version": "0.1.8",
      "runtime_loaded_from_requested_source": true,
      "fixture": {
        "character": "A",
        "target": "tool_selection",
        "characters": 15000,
        "marker_prefix_characters": 24,
        "payload_character_repetitions": 14862,
        "frozen_monotonic": 100.0,
        "max_steps": 2,
        "timeout_seconds": 120
      },
      "projected_context_sizes": {
        "context_projection": {
          "characters": 14964,
          "utf8_bytes": 14964
        },
        "tool_selection": {
          "characters": 15000,
          "utf8_bytes": 15000
        }
      },
      "expected": {
        "status": "done",
        "scripted_decisions": 1,
        "effects": 1,
        "intent_events": 1,
        "reason": null,
        "context_budget": null
      },
      "result": {
        "status": "done",
        "reason": null,
        "context_budget": null,
        "failure_metadata_correct": true,
        "success_metadata_correct": false,
        "behavior_correct": true,
        "scripted_decisions": 1,
        "effects": 1,
        "intent_events": 1,
        "native_inference_attempts": 0,
        "request_context_sizes": [
          {
            "stage": "tool_selection",
            "characters": 15000,
            "utf8_bytes": 15000
          }
        ],
        "returned_call_telemetry": [
          {
            "stage": "tool_selection",
            "context_characters": null,
            "context_character_limit": null,
            "context_bytes": 15000
          }
        ],
        "terminal_marker_exposures": 0,
        "reported_scripted_usage_complete": true
      },
      "correct": false,
      "whole_operation_seconds": 0.0017804160015657544,
      "returned_context_bytes": 403,
      "actual_model_usage": {
        "requests": 0,
        "models": {},
        "input_tokens": 0,
        "output_tokens": 0,
        "usage_complete": true
      },
      "arm": "baseline",
      "repeat": 1
    },
    {
      "case": "ascii_tool_exact",
      "runtime_version": "0.1.9",
      "runtime_loaded_from_requested_source": true,
      "fixture": {
        "character": "A",
        "target": "tool_selection",
        "characters": 15000,
        "marker_prefix_characters": 24,
        "payload_character_repetitions": 14862,
        "frozen_monotonic": 100.0,
        "max_steps": 2,
        "timeout_seconds": 120
      },
      "projected_context_sizes": {
        "context_projection": {
          "characters": 14964,
          "utf8_bytes": 14964
        },
        "tool_selection": {
          "characters": 15000,
          "utf8_bytes": 15000
        }
      },
      "expected": {
        "status": "done",
        "scripted_decisions": 1,
        "effects": 1,
        "intent_events": 1,
        "reason": null,
        "context_budget": null
      },
      "result": {
        "status": "done",
        "reason": null,
        "context_budget": null,
        "failure_metadata_correct": true,
        "success_metadata_correct": true,
        "behavior_correct": true,
        "scripted_decisions": 1,
        "effects": 1,
        "intent_events": 1,
        "native_inference_attempts": 0,
        "request_context_sizes": [
          {
            "stage": "tool_selection",
            "characters": 15000,
            "utf8_bytes": 15000
          }
        ],
        "returned_call_telemetry": [
          {
            "stage": "tool_selection",
            "context_characters": 15000,
            "context_character_limit": 15000,
            "context_bytes": 15000
          }
        ],
        "terminal_marker_exposures": 0,
        "reported_scripted_usage_complete": true
      },
      "correct": true,
      "whole_operation_seconds": 0.0018522080063121393,
      "returned_context_bytes": 466,
      "actual_model_usage": {
        "requests": 0,
        "models": {},
        "input_tokens": 0,
        "output_tokens": 0,
        "usage_complete": true
      },
      "arm": "candidate",
      "repeat": 1
    },
    {
      "case": "ascii_tool_exact",
      "runtime_version": "0.1.9",
      "runtime_loaded_from_requested_source": true,
      "fixture": {
        "character": "A",
        "target": "tool_selection",
        "characters": 15000,
        "marker_prefix_characters": 24,
        "payload_character_repetitions": 14862,
        "frozen_monotonic": 100.0,
        "max_steps": 2,
        "timeout_seconds": 120
      },
      "projected_context_sizes": {
        "context_projection": {
          "characters": 14964,
          "utf8_bytes": 14964
        },
        "tool_selection": {
          "characters": 15000,
          "utf8_bytes": 15000
        }
      },
      "expected": {
        "status": "done",
        "scripted_decisions": 1,
        "effects": 1,
        "intent_events": 1,
        "reason": null,
        "context_budget": null
      },
      "result": {
        "status": "done",
        "reason": null,
        "context_budget": null,
        "failure_metadata_correct": true,
        "success_metadata_correct": true,
        "behavior_correct": true,
        "scripted_decisions": 1,
        "effects": 1,
        "intent_events": 1,
        "native_inference_attempts": 0,
        "request_context_sizes": [
          {
            "stage": "tool_selection",
            "characters": 15000,
            "utf8_bytes": 15000
          }
        ],
        "returned_call_telemetry": [
          {
            "stage": "tool_selection",
            "context_characters": 15000,
            "context_character_limit": 15000,
            "context_bytes": 15000
          }
        ],
        "terminal_marker_exposures": 0,
        "reported_scripted_usage_complete": true
      },
      "correct": true,
      "whole_operation_seconds": 0.001707834002445452,
      "returned_context_bytes": 466,
      "actual_model_usage": {
        "requests": 0,
        "models": {},
        "input_tokens": 0,
        "output_tokens": 0,
        "usage_complete": true
      },
      "arm": "candidate",
      "repeat": 2
    },
    {
      "case": "ascii_tool_exact",
      "runtime_version": "0.1.8",
      "runtime_loaded_from_requested_source": true,
      "fixture": {
        "character": "A",
        "target": "tool_selection",
        "characters": 15000,
        "marker_prefix_characters": 24,
        "payload_character_repetitions": 14862,
        "frozen_monotonic": 100.0,
        "max_steps": 2,
        "timeout_seconds": 120
      },
      "projected_context_sizes": {
        "context_projection": {
          "characters": 14964,
          "utf8_bytes": 14964
        },
        "tool_selection": {
          "characters": 15000,
          "utf8_bytes": 15000
        }
      },
      "expected": {
        "status": "done",
        "scripted_decisions": 1,
        "effects": 1,
        "intent_events": 1,
        "reason": null,
        "context_budget": null
      },
      "result": {
        "status": "done",
        "reason": null,
        "context_budget": null,
        "failure_metadata_correct": true,
        "success_metadata_correct": false,
        "behavior_correct": true,
        "scripted_decisions": 1,
        "effects": 1,
        "intent_events": 1,
        "native_inference_attempts": 0,
        "request_context_sizes": [
          {
            "stage": "tool_selection",
            "characters": 15000,
            "utf8_bytes": 15000
          }
        ],
        "returned_call_telemetry": [
          {
            "stage": "tool_selection",
            "context_characters": null,
            "context_character_limit": null,
            "context_bytes": 15000
          }
        ],
        "terminal_marker_exposures": 0,
        "reported_scripted_usage_complete": true
      },
      "correct": false,
      "whole_operation_seconds": 0.0019281249988125637,
      "returned_context_bytes": 403,
      "actual_model_usage": {
        "requests": 0,
        "models": {},
        "input_tokens": 0,
        "output_tokens": 0,
        "usage_complete": true
      },
      "arm": "baseline",
      "repeat": 2
    },
    {
      "case": "ascii_tool_exact",
      "runtime_version": "0.1.8",
      "runtime_loaded_from_requested_source": true,
      "fixture": {
        "character": "A",
        "target": "tool_selection",
        "characters": 15000,
        "marker_prefix_characters": 24,
        "payload_character_repetitions": 14862,
        "frozen_monotonic": 100.0,
        "max_steps": 2,
        "timeout_seconds": 120
      },
      "projected_context_sizes": {
        "context_projection": {
          "characters": 14964,
          "utf8_bytes": 14964
        },
        "tool_selection": {
          "characters": 15000,
          "utf8_bytes": 15000
        }
      },
      "expected": {
        "status": "done",
        "scripted_decisions": 1,
        "effects": 1,
        "intent_events": 1,
        "reason": null,
        "context_budget": null
      },
      "result": {
        "status": "done",
        "reason": null,
        "context_budget": null,
        "failure_metadata_correct": true,
        "success_metadata_correct": false,
        "behavior_correct": true,
        "scripted_decisions": 1,
        "effects": 1,
        "intent_events": 1,
        "native_inference_attempts": 0,
        "request_context_sizes": [
          {
            "stage": "tool_selection",
            "characters": 15000,
            "utf8_bytes": 15000
          }
        ],
        "returned_call_telemetry": [
          {
            "stage": "tool_selection",
            "context_characters": null,
            "context_character_limit": null,
            "context_bytes": 15000
          }
        ],
        "terminal_marker_exposures": 0,
        "reported_scripted_usage_complete": true
      },
      "correct": false,
      "whole_operation_seconds": 0.00189124999451451,
      "returned_context_bytes": 403,
      "actual_model_usage": {
        "requests": 0,
        "models": {},
        "input_tokens": 0,
        "output_tokens": 0,
        "usage_complete": true
      },
      "arm": "baseline",
      "repeat": 3
    },
    {
      "case": "ascii_tool_exact",
      "runtime_version": "0.1.9",
      "runtime_loaded_from_requested_source": true,
      "fixture": {
        "character": "A",
        "target": "tool_selection",
        "characters": 15000,
        "marker_prefix_characters": 24,
        "payload_character_repetitions": 14862,
        "frozen_monotonic": 100.0,
        "max_steps": 2,
        "timeout_seconds": 120
      },
      "projected_context_sizes": {
        "context_projection": {
          "characters": 14964,
          "utf8_bytes": 14964
        },
        "tool_selection": {
          "characters": 15000,
          "utf8_bytes": 15000
        }
      },
      "expected": {
        "status": "done",
        "scripted_decisions": 1,
        "effects": 1,
        "intent_events": 1,
        "reason": null,
        "context_budget": null
      },
      "result": {
        "status": "done",
        "reason": null,
        "context_budget": null,
        "failure_metadata_correct": true,
        "success_metadata_correct": true,
        "behavior_correct": true,
        "scripted_decisions": 1,
        "effects": 1,
        "intent_events": 1,
        "native_inference_attempts": 0,
        "request_context_sizes": [
          {
            "stage": "tool_selection",
            "characters": 15000,
            "utf8_bytes": 15000
          }
        ],
        "returned_call_telemetry": [
          {
            "stage": "tool_selection",
            "context_characters": 15000,
            "context_character_limit": 15000,
            "context_bytes": 15000
          }
        ],
        "terminal_marker_exposures": 0,
        "reported_scripted_usage_complete": true
      },
      "correct": true,
      "whole_operation_seconds": 0.0016662080015521497,
      "returned_context_bytes": 466,
      "actual_model_usage": {
        "requests": 0,
        "models": {},
        "input_tokens": 0,
        "output_tokens": 0,
        "usage_complete": true
      },
      "arm": "candidate",
      "repeat": 3
    },
    {
      "case": "ascii_tool_under",
      "runtime_version": "0.1.8",
      "runtime_loaded_from_requested_source": true,
      "fixture": {
        "character": "A",
        "target": "tool_selection",
        "characters": 14999,
        "marker_prefix_characters": 24,
        "payload_character_repetitions": 14861,
        "frozen_monotonic": 100.0,
        "max_steps": 2,
        "timeout_seconds": 120
      },
      "projected_context_sizes": {
        "context_projection": {
          "characters": 14963,
          "utf8_bytes": 14963
        },
        "tool_selection": {
          "characters": 14999,
          "utf8_bytes": 14999
        }
      },
      "expected": {
        "status": "done",
        "scripted_decisions": 1,
        "effects": 1,
        "intent_events": 1,
        "reason": null,
        "context_budget": null
      },
      "result": {
        "status": "done",
        "reason": null,
        "context_budget": null,
        "failure_metadata_correct": true,
        "success_metadata_correct": false,
        "behavior_correct": true,
        "scripted_decisions": 1,
        "effects": 1,
        "intent_events": 1,
        "native_inference_attempts": 0,
        "request_context_sizes": [
          {
            "stage": "tool_selection",
            "characters": 14999,
            "utf8_bytes": 14999
          }
        ],
        "returned_call_telemetry": [
          {
            "stage": "tool_selection",
            "context_characters": null,
            "context_character_limit": null,
            "context_bytes": 14999
          }
        ],
        "terminal_marker_exposures": 0,
        "reported_scripted_usage_complete": true
      },
      "correct": false,
      "whole_operation_seconds": 0.0017345840024063364,
      "returned_context_bytes": 403,
      "actual_model_usage": {
        "requests": 0,
        "models": {},
        "input_tokens": 0,
        "output_tokens": 0,
        "usage_complete": true
      },
      "arm": "baseline",
      "repeat": 1
    },
    {
      "case": "ascii_tool_under",
      "runtime_version": "0.1.9",
      "runtime_loaded_from_requested_source": true,
      "fixture": {
        "character": "A",
        "target": "tool_selection",
        "characters": 14999,
        "marker_prefix_characters": 24,
        "payload_character_repetitions": 14861,
        "frozen_monotonic": 100.0,
        "max_steps": 2,
        "timeout_seconds": 120
      },
      "projected_context_sizes": {
        "context_projection": {
          "characters": 14963,
          "utf8_bytes": 14963
        },
        "tool_selection": {
          "characters": 14999,
          "utf8_bytes": 14999
        }
      },
      "expected": {
        "status": "done",
        "scripted_decisions": 1,
        "effects": 1,
        "intent_events": 1,
        "reason": null,
        "context_budget": null
      },
      "result": {
        "status": "done",
        "reason": null,
        "context_budget": null,
        "failure_metadata_correct": true,
        "success_metadata_correct": true,
        "behavior_correct": true,
        "scripted_decisions": 1,
        "effects": 1,
        "intent_events": 1,
        "native_inference_attempts": 0,
        "request_context_sizes": [
          {
            "stage": "tool_selection",
            "characters": 14999,
            "utf8_bytes": 14999
          }
        ],
        "returned_call_telemetry": [
          {
            "stage": "tool_selection",
            "context_characters": 14999,
            "context_character_limit": 15000,
            "context_bytes": 14999
          }
        ],
        "terminal_marker_exposures": 0,
        "reported_scripted_usage_complete": true
      },
      "correct": true,
      "whole_operation_seconds": 0.0022309589985525236,
      "returned_context_bytes": 466,
      "actual_model_usage": {
        "requests": 0,
        "models": {},
        "input_tokens": 0,
        "output_tokens": 0,
        "usage_complete": true
      },
      "arm": "candidate",
      "repeat": 1
    },
    {
      "case": "ascii_tool_under",
      "runtime_version": "0.1.9",
      "runtime_loaded_from_requested_source": true,
      "fixture": {
        "character": "A",
        "target": "tool_selection",
        "characters": 14999,
        "marker_prefix_characters": 24,
        "payload_character_repetitions": 14861,
        "frozen_monotonic": 100.0,
        "max_steps": 2,
        "timeout_seconds": 120
      },
      "projected_context_sizes": {
        "context_projection": {
          "characters": 14963,
          "utf8_bytes": 14963
        },
        "tool_selection": {
          "characters": 14999,
          "utf8_bytes": 14999
        }
      },
      "expected": {
        "status": "done",
        "scripted_decisions": 1,
        "effects": 1,
        "intent_events": 1,
        "reason": null,
        "context_budget": null
      },
      "result": {
        "status": "done",
        "reason": null,
        "context_budget": null,
        "failure_metadata_correct": true,
        "success_metadata_correct": true,
        "behavior_correct": true,
        "scripted_decisions": 1,
        "effects": 1,
        "intent_events": 1,
        "native_inference_attempts": 0,
        "request_context_sizes": [
          {
            "stage": "tool_selection",
            "characters": 14999,
            "utf8_bytes": 14999
          }
        ],
        "returned_call_telemetry": [
          {
            "stage": "tool_selection",
            "context_characters": 14999,
            "context_character_limit": 15000,
            "context_bytes": 14999
          }
        ],
        "terminal_marker_exposures": 0,
        "reported_scripted_usage_complete": true
      },
      "correct": true,
      "whole_operation_seconds": 0.0017284580098930746,
      "returned_context_bytes": 466,
      "actual_model_usage": {
        "requests": 0,
        "models": {},
        "input_tokens": 0,
        "output_tokens": 0,
        "usage_complete": true
      },
      "arm": "candidate",
      "repeat": 2
    },
    {
      "case": "ascii_tool_under",
      "runtime_version": "0.1.8",
      "runtime_loaded_from_requested_source": true,
      "fixture": {
        "character": "A",
        "target": "tool_selection",
        "characters": 14999,
        "marker_prefix_characters": 24,
        "payload_character_repetitions": 14861,
        "frozen_monotonic": 100.0,
        "max_steps": 2,
        "timeout_seconds": 120
      },
      "projected_context_sizes": {
        "context_projection": {
          "characters": 14963,
          "utf8_bytes": 14963
        },
        "tool_selection": {
          "characters": 14999,
          "utf8_bytes": 14999
        }
      },
      "expected": {
        "status": "done",
        "scripted_decisions": 1,
        "effects": 1,
        "intent_events": 1,
        "reason": null,
        "context_budget": null
      },
      "result": {
        "status": "done",
        "reason": null,
        "context_budget": null,
        "failure_metadata_correct": true,
        "success_metadata_correct": false,
        "behavior_correct": true,
        "scripted_decisions": 1,
        "effects": 1,
        "intent_events": 1,
        "native_inference_attempts": 0,
        "request_context_sizes": [
          {
            "stage": "tool_selection",
            "characters": 14999,
            "utf8_bytes": 14999
          }
        ],
        "returned_call_telemetry": [
          {
            "stage": "tool_selection",
            "context_characters": null,
            "context_character_limit": null,
            "context_bytes": 14999
          }
        ],
        "terminal_marker_exposures": 0,
        "reported_scripted_usage_complete": true
      },
      "correct": false,
      "whole_operation_seconds": 0.0017714999994495884,
      "returned_context_bytes": 403,
      "actual_model_usage": {
        "requests": 0,
        "models": {},
        "input_tokens": 0,
        "output_tokens": 0,
        "usage_complete": true
      },
      "arm": "baseline",
      "repeat": 2
    },
    {
      "case": "ascii_tool_under",
      "runtime_version": "0.1.8",
      "runtime_loaded_from_requested_source": true,
      "fixture": {
        "character": "A",
        "target": "tool_selection",
        "characters": 14999,
        "marker_prefix_characters": 24,
        "payload_character_repetitions": 14861,
        "frozen_monotonic": 100.0,
        "max_steps": 2,
        "timeout_seconds": 120
      },
      "projected_context_sizes": {
        "context_projection": {
          "characters": 14963,
          "utf8_bytes": 14963
        },
        "tool_selection": {
          "characters": 14999,
          "utf8_bytes": 14999
        }
      },
      "expected": {
        "status": "done",
        "scripted_decisions": 1,
        "effects": 1,
        "intent_events": 1,
        "reason": null,
        "context_budget": null
      },
      "result": {
        "status": "done",
        "reason": null,
        "context_budget": null,
        "failure_metadata_correct": true,
        "success_metadata_correct": false,
        "behavior_correct": true,
        "scripted_decisions": 1,
        "effects": 1,
        "intent_events": 1,
        "native_inference_attempts": 0,
        "request_context_sizes": [
          {
            "stage": "tool_selection",
            "characters": 14999,
            "utf8_bytes": 14999
          }
        ],
        "returned_call_telemetry": [
          {
            "stage": "tool_selection",
            "context_characters": null,
            "context_character_limit": null,
            "context_bytes": 14999
          }
        ],
        "terminal_marker_exposures": 0,
        "reported_scripted_usage_complete": true
      },
      "correct": false,
      "whole_operation_seconds": 0.0016563340032007545,
      "returned_context_bytes": 403,
      "actual_model_usage": {
        "requests": 0,
        "models": {},
        "input_tokens": 0,
        "output_tokens": 0,
        "usage_complete": true
      },
      "arm": "baseline",
      "repeat": 3
    },
    {
      "case": "ascii_tool_under",
      "runtime_version": "0.1.9",
      "runtime_loaded_from_requested_source": true,
      "fixture": {
        "character": "A",
        "target": "tool_selection",
        "characters": 14999,
        "marker_prefix_characters": 24,
        "payload_character_repetitions": 14861,
        "frozen_monotonic": 100.0,
        "max_steps": 2,
        "timeout_seconds": 120
      },
      "projected_context_sizes": {
        "context_projection": {
          "characters": 14963,
          "utf8_bytes": 14963
        },
        "tool_selection": {
          "characters": 14999,
          "utf8_bytes": 14999
        }
      },
      "expected": {
        "status": "done",
        "scripted_decisions": 1,
        "effects": 1,
        "intent_events": 1,
        "reason": null,
        "context_budget": null
      },
      "result": {
        "status": "done",
        "reason": null,
        "context_budget": null,
        "failure_metadata_correct": true,
        "success_metadata_correct": true,
        "behavior_correct": true,
        "scripted_decisions": 1,
        "effects": 1,
        "intent_events": 1,
        "native_inference_attempts": 0,
        "request_context_sizes": [
          {
            "stage": "tool_selection",
            "characters": 14999,
            "utf8_bytes": 14999
          }
        ],
        "returned_call_telemetry": [
          {
            "stage": "tool_selection",
            "context_characters": 14999,
            "context_character_limit": 15000,
            "context_bytes": 14999
          }
        ],
        "terminal_marker_exposures": 0,
        "reported_scripted_usage_complete": true
      },
      "correct": true,
      "whole_operation_seconds": 0.0018994590063812211,
      "returned_context_bytes": 466,
      "actual_model_usage": {
        "requests": 0,
        "models": {},
        "input_tokens": 0,
        "output_tokens": 0,
        "usage_complete": true
      },
      "arm": "candidate",
      "repeat": 3
    },
    {
      "case": "chinese_tool_exact",
      "runtime_version": "0.1.8",
      "runtime_loaded_from_requested_source": true,
      "fixture": {
        "character": "\u4e2d",
        "target": "tool_selection",
        "characters": 15000,
        "marker_prefix_characters": 24,
        "payload_character_repetitions": 14862,
        "frozen_monotonic": 100.0,
        "max_steps": 2,
        "timeout_seconds": 120
      },
      "projected_context_sizes": {
        "context_projection": {
          "characters": 14964,
          "utf8_bytes": 44688
        },
        "tool_selection": {
          "characters": 15000,
          "utf8_bytes": 44724
        }
      },
      "expected": {
        "status": "done",
        "scripted_decisions": 1,
        "effects": 1,
        "intent_events": 1,
        "reason": null,
        "context_budget": null
      },
      "result": {
        "status": "done",
        "reason": null,
        "context_budget": null,
        "failure_metadata_correct": true,
        "success_metadata_correct": false,
        "behavior_correct": true,
        "scripted_decisions": 1,
        "effects": 1,
        "intent_events": 1,
        "native_inference_attempts": 0,
        "request_context_sizes": [
          {
            "stage": "tool_selection",
            "characters": 15000,
            "utf8_bytes": 44724
          }
        ],
        "returned_call_telemetry": [
          {
            "stage": "tool_selection",
            "context_characters": null,
            "context_character_limit": null,
            "context_bytes": 44724
          }
        ],
        "terminal_marker_exposures": 0,
        "reported_scripted_usage_complete": true
      },
      "correct": false,
      "whole_operation_seconds": 0.0019604590052040294,
      "returned_context_bytes": 403,
      "actual_model_usage": {
        "requests": 0,
        "models": {},
        "input_tokens": 0,
        "output_tokens": 0,
        "usage_complete": true
      },
      "arm": "baseline",
      "repeat": 1
    },
    {
      "case": "chinese_tool_exact",
      "runtime_version": "0.1.9",
      "runtime_loaded_from_requested_source": true,
      "fixture": {
        "character": "\u4e2d",
        "target": "tool_selection",
        "characters": 15000,
        "marker_prefix_characters": 24,
        "payload_character_repetitions": 14862,
        "frozen_monotonic": 100.0,
        "max_steps": 2,
        "timeout_seconds": 120
      },
      "projected_context_sizes": {
        "context_projection": {
          "characters": 14964,
          "utf8_bytes": 44688
        },
        "tool_selection": {
          "characters": 15000,
          "utf8_bytes": 44724
        }
      },
      "expected": {
        "status": "done",
        "scripted_decisions": 1,
        "effects": 1,
        "intent_events": 1,
        "reason": null,
        "context_budget": null
      },
      "result": {
        "status": "done",
        "reason": null,
        "context_budget": null,
        "failure_metadata_correct": true,
        "success_metadata_correct": true,
        "behavior_correct": true,
        "scripted_decisions": 1,
        "effects": 1,
        "intent_events": 1,
        "native_inference_attempts": 0,
        "request_context_sizes": [
          {
            "stage": "tool_selection",
            "characters": 15000,
            "utf8_bytes": 44724
          }
        ],
        "returned_call_telemetry": [
          {
            "stage": "tool_selection",
            "context_characters": 15000,
            "context_character_limit": 15000,
            "context_bytes": 44724
          }
        ],
        "terminal_marker_exposures": 0,
        "reported_scripted_usage_complete": true
      },
      "correct": true,
      "whole_operation_seconds": 0.0017920000100275502,
      "returned_context_bytes": 466,
      "actual_model_usage": {
        "requests": 0,
        "models": {},
        "input_tokens": 0,
        "output_tokens": 0,
        "usage_complete": true
      },
      "arm": "candidate",
      "repeat": 1
    },
    {
      "case": "chinese_tool_exact",
      "runtime_version": "0.1.9",
      "runtime_loaded_from_requested_source": true,
      "fixture": {
        "character": "\u4e2d",
        "target": "tool_selection",
        "characters": 15000,
        "marker_prefix_characters": 24,
        "payload_character_repetitions": 14862,
        "frozen_monotonic": 100.0,
        "max_steps": 2,
        "timeout_seconds": 120
      },
      "projected_context_sizes": {
        "context_projection": {
          "characters": 14964,
          "utf8_bytes": 44688
        },
        "tool_selection": {
          "characters": 15000,
          "utf8_bytes": 44724
        }
      },
      "expected": {
        "status": "done",
        "scripted_decisions": 1,
        "effects": 1,
        "intent_events": 1,
        "reason": null,
        "context_budget": null
      },
      "result": {
        "status": "done",
        "reason": null,
        "context_budget": null,
        "failure_metadata_correct": true,
        "success_metadata_correct": true,
        "behavior_correct": true,
        "scripted_decisions": 1,
        "effects": 1,
        "intent_events": 1,
        "native_inference_attempts": 0,
        "request_context_sizes": [
          {
            "stage": "tool_selection",
            "characters": 15000,
            "utf8_bytes": 44724
          }
        ],
        "returned_call_telemetry": [
          {
            "stage": "tool_selection",
            "context_characters": 15000,
            "context_character_limit": 15000,
            "context_bytes": 44724
          }
        ],
        "terminal_marker_exposures": 0,
        "reported_scripted_usage_complete": true
      },
      "correct": true,
      "whole_operation_seconds": 0.0018523750040913,
      "returned_context_bytes": 466,
      "actual_model_usage": {
        "requests": 0,
        "models": {},
        "input_tokens": 0,
        "output_tokens": 0,
        "usage_complete": true
      },
      "arm": "candidate",
      "repeat": 2
    },
    {
      "case": "chinese_tool_exact",
      "runtime_version": "0.1.8",
      "runtime_loaded_from_requested_source": true,
      "fixture": {
        "character": "\u4e2d",
        "target": "tool_selection",
        "characters": 15000,
        "marker_prefix_characters": 24,
        "payload_character_repetitions": 14862,
        "frozen_monotonic": 100.0,
        "max_steps": 2,
        "timeout_seconds": 120
      },
      "projected_context_sizes": {
        "context_projection": {
          "characters": 14964,
          "utf8_bytes": 44688
        },
        "tool_selection": {
          "characters": 15000,
          "utf8_bytes": 44724
        }
      },
      "expected": {
        "status": "done",
        "scripted_decisions": 1,
        "effects": 1,
        "intent_events": 1,
        "reason": null,
        "context_budget": null
      },
      "result": {
        "status": "done",
        "reason": null,
        "context_budget": null,
        "failure_metadata_correct": true,
        "success_metadata_correct": false,
        "behavior_correct": true,
        "scripted_decisions": 1,
        "effects": 1,
        "intent_events": 1,
        "native_inference_attempts": 0,
        "request_context_sizes": [
          {
            "stage": "tool_selection",
            "characters": 15000,
            "utf8_bytes": 44724
          }
        ],
        "returned_call_telemetry": [
          {
            "stage": "tool_selection",
            "context_characters": null,
            "context_character_limit": null,
            "context_bytes": 44724
          }
        ],
        "terminal_marker_exposures": 0,
        "reported_scripted_usage_complete": true
      },
      "correct": false,
      "whole_operation_seconds": 0.0017651660018600523,
      "returned_context_bytes": 403,
      "actual_model_usage": {
        "requests": 0,
        "models": {},
        "input_tokens": 0,
        "output_tokens": 0,
        "usage_complete": true
      },
      "arm": "baseline",
      "repeat": 2
    },
    {
      "case": "chinese_tool_exact",
      "runtime_version": "0.1.8",
      "runtime_loaded_from_requested_source": true,
      "fixture": {
        "character": "\u4e2d",
        "target": "tool_selection",
        "characters": 15000,
        "marker_prefix_characters": 24,
        "payload_character_repetitions": 14862,
        "frozen_monotonic": 100.0,
        "max_steps": 2,
        "timeout_seconds": 120
      },
      "projected_context_sizes": {
        "context_projection": {
          "characters": 14964,
          "utf8_bytes": 44688
        },
        "tool_selection": {
          "characters": 15000,
          "utf8_bytes": 44724
        }
      },
      "expected": {
        "status": "done",
        "scripted_decisions": 1,
        "effects": 1,
        "intent_events": 1,
        "reason": null,
        "context_budget": null
      },
      "result": {
        "status": "done",
        "reason": null,
        "context_budget": null,
        "failure_metadata_correct": true,
        "success_metadata_correct": false,
        "behavior_correct": true,
        "scripted_decisions": 1,
        "effects": 1,
        "intent_events": 1,
        "native_inference_attempts": 0,
        "request_context_sizes": [
          {
            "stage": "tool_selection",
            "characters": 15000,
            "utf8_bytes": 44724
          }
        ],
        "returned_call_telemetry": [
          {
            "stage": "tool_selection",
            "context_characters": null,
            "context_character_limit": null,
            "context_bytes": 44724
          }
        ],
        "terminal_marker_exposures": 0,
        "reported_scripted_usage_complete": true
      },
      "correct": false,
      "whole_operation_seconds": 0.0017382079968228936,
      "returned_context_bytes": 403,
      "actual_model_usage": {
        "requests": 0,
        "models": {},
        "input_tokens": 0,
        "output_tokens": 0,
        "usage_complete": true
      },
      "arm": "baseline",
      "repeat": 3
    },
    {
      "case": "chinese_tool_exact",
      "runtime_version": "0.1.9",
      "runtime_loaded_from_requested_source": true,
      "fixture": {
        "character": "\u4e2d",
        "target": "tool_selection",
        "characters": 15000,
        "marker_prefix_characters": 24,
        "payload_character_repetitions": 14862,
        "frozen_monotonic": 100.0,
        "max_steps": 2,
        "timeout_seconds": 120
      },
      "projected_context_sizes": {
        "context_projection": {
          "characters": 14964,
          "utf8_bytes": 44688
        },
        "tool_selection": {
          "characters": 15000,
          "utf8_bytes": 44724
        }
      },
      "expected": {
        "status": "done",
        "scripted_decisions": 1,
        "effects": 1,
        "intent_events": 1,
        "reason": null,
        "context_budget": null
      },
      "result": {
        "status": "done",
        "reason": null,
        "context_budget": null,
        "failure_metadata_correct": true,
        "success_metadata_correct": true,
        "behavior_correct": true,
        "scripted_decisions": 1,
        "effects": 1,
        "intent_events": 1,
        "native_inference_attempts": 0,
        "request_context_sizes": [
          {
            "stage": "tool_selection",
            "characters": 15000,
            "utf8_bytes": 44724
          }
        ],
        "returned_call_telemetry": [
          {
            "stage": "tool_selection",
            "context_characters": 15000,
            "context_character_limit": 15000,
            "context_bytes": 44724
          }
        ],
        "terminal_marker_exposures": 0,
        "reported_scripted_usage_complete": true
      },
      "correct": true,
      "whole_operation_seconds": 0.0019926670065615326,
      "returned_context_bytes": 466,
      "actual_model_usage": {
        "requests": 0,
        "models": {},
        "input_tokens": 0,
        "output_tokens": 0,
        "usage_complete": true
      },
      "arm": "candidate",
      "repeat": 3
    },
    {
      "case": "emoji_tool_exact",
      "runtime_version": "0.1.8",
      "runtime_loaded_from_requested_source": true,
      "fixture": {
        "character": "\ud83d\ude42",
        "target": "tool_selection",
        "characters": 15000,
        "marker_prefix_characters": 24,
        "payload_character_repetitions": 14862,
        "frozen_monotonic": 100.0,
        "max_steps": 2,
        "timeout_seconds": 120
      },
      "projected_context_sizes": {
        "context_projection": {
          "characters": 14964,
          "utf8_bytes": 59550
        },
        "tool_selection": {
          "characters": 15000,
          "utf8_bytes": 59586
        }
      },
      "expected": {
        "status": "done",
        "scripted_decisions": 1,
        "effects": 1,
        "intent_events": 1,
        "reason": null,
        "context_budget": null
      },
      "result": {
        "status": "done",
        "reason": null,
        "context_budget": null,
        "failure_metadata_correct": true,
        "success_metadata_correct": false,
        "behavior_correct": true,
        "scripted_decisions": 1,
        "effects": 1,
        "intent_events": 1,
        "native_inference_attempts": 0,
        "request_context_sizes": [
          {
            "stage": "tool_selection",
            "characters": 15000,
            "utf8_bytes": 59586
          }
        ],
        "returned_call_telemetry": [
          {
            "stage": "tool_selection",
            "context_characters": null,
            "context_character_limit": null,
            "context_bytes": 59586
          }
        ],
        "terminal_marker_exposures": 0,
        "reported_scripted_usage_complete": true
      },
      "correct": false,
      "whole_operation_seconds": 0.002033541997661814,
      "returned_context_bytes": 403,
      "actual_model_usage": {
        "requests": 0,
        "models": {},
        "input_tokens": 0,
        "output_tokens": 0,
        "usage_complete": true
      },
      "arm": "baseline",
      "repeat": 1
    },
    {
      "case": "emoji_tool_exact",
      "runtime_version": "0.1.9",
      "runtime_loaded_from_requested_source": true,
      "fixture": {
        "character": "\ud83d\ude42",
        "target": "tool_selection",
        "characters": 15000,
        "marker_prefix_characters": 24,
        "payload_character_repetitions": 14862,
        "frozen_monotonic": 100.0,
        "max_steps": 2,
        "timeout_seconds": 120
      },
      "projected_context_sizes": {
        "context_projection": {
          "characters": 14964,
          "utf8_bytes": 59550
        },
        "tool_selection": {
          "characters": 15000,
          "utf8_bytes": 59586
        }
      },
      "expected": {
        "status": "done",
        "scripted_decisions": 1,
        "effects": 1,
        "intent_events": 1,
        "reason": null,
        "context_budget": null
      },
      "result": {
        "status": "done",
        "reason": null,
        "context_budget": null,
        "failure_metadata_correct": true,
        "success_metadata_correct": true,
        "behavior_correct": true,
        "scripted_decisions": 1,
        "effects": 1,
        "intent_events": 1,
        "native_inference_attempts": 0,
        "request_context_sizes": [
          {
            "stage": "tool_selection",
            "characters": 15000,
            "utf8_bytes": 59586
          }
        ],
        "returned_call_telemetry": [
          {
            "stage": "tool_selection",
            "context_characters": 15000,
            "context_character_limit": 15000,
            "context_bytes": 59586
          }
        ],
        "terminal_marker_exposures": 0,
        "reported_scripted_usage_complete": true
      },
      "correct": true,
      "whole_operation_seconds": 0.0022809580113971606,
      "returned_context_bytes": 466,
      "actual_model_usage": {
        "requests": 0,
        "models": {},
        "input_tokens": 0,
        "output_tokens": 0,
        "usage_complete": true
      },
      "arm": "candidate",
      "repeat": 1
    },
    {
      "case": "emoji_tool_exact",
      "runtime_version": "0.1.9",
      "runtime_loaded_from_requested_source": true,
      "fixture": {
        "character": "\ud83d\ude42",
        "target": "tool_selection",
        "characters": 15000,
        "marker_prefix_characters": 24,
        "payload_character_repetitions": 14862,
        "frozen_monotonic": 100.0,
        "max_steps": 2,
        "timeout_seconds": 120
      },
      "projected_context_sizes": {
        "context_projection": {
          "characters": 14964,
          "utf8_bytes": 59550
        },
        "tool_selection": {
          "characters": 15000,
          "utf8_bytes": 59586
        }
      },
      "expected": {
        "status": "done",
        "scripted_decisions": 1,
        "effects": 1,
        "intent_events": 1,
        "reason": null,
        "context_budget": null
      },
      "result": {
        "status": "done",
        "reason": null,
        "context_budget": null,
        "failure_metadata_correct": true,
        "success_metadata_correct": true,
        "behavior_correct": true,
        "scripted_decisions": 1,
        "effects": 1,
        "intent_events": 1,
        "native_inference_attempts": 0,
        "request_context_sizes": [
          {
            "stage": "tool_selection",
            "characters": 15000,
            "utf8_bytes": 59586
          }
        ],
        "returned_call_telemetry": [
          {
            "stage": "tool_selection",
            "context_characters": 15000,
            "context_character_limit": 15000,
            "context_bytes": 59586
          }
        ],
        "terminal_marker_exposures": 0,
        "reported_scripted_usage_complete": true
      },
      "correct": true,
      "whole_operation_seconds": 0.0019386659987503663,
      "returned_context_bytes": 466,
      "actual_model_usage": {
        "requests": 0,
        "models": {},
        "input_tokens": 0,
        "output_tokens": 0,
        "usage_complete": true
      },
      "arm": "candidate",
      "repeat": 2
    },
    {
      "case": "emoji_tool_exact",
      "runtime_version": "0.1.8",
      "runtime_loaded_from_requested_source": true,
      "fixture": {
        "character": "\ud83d\ude42",
        "target": "tool_selection",
        "characters": 15000,
        "marker_prefix_characters": 24,
        "payload_character_repetitions": 14862,
        "frozen_monotonic": 100.0,
        "max_steps": 2,
        "timeout_seconds": 120
      },
      "projected_context_sizes": {
        "context_projection": {
          "characters": 14964,
          "utf8_bytes": 59550
        },
        "tool_selection": {
          "characters": 15000,
          "utf8_bytes": 59586
        }
      },
      "expected": {
        "status": "done",
        "scripted_decisions": 1,
        "effects": 1,
        "intent_events": 1,
        "reason": null,
        "context_budget": null
      },
      "result": {
        "status": "done",
        "reason": null,
        "context_budget": null,
        "failure_metadata_correct": true,
        "success_metadata_correct": false,
        "behavior_correct": true,
        "scripted_decisions": 1,
        "effects": 1,
        "intent_events": 1,
        "native_inference_attempts": 0,
        "request_context_sizes": [
          {
            "stage": "tool_selection",
            "characters": 15000,
            "utf8_bytes": 59586
          }
        ],
        "returned_call_telemetry": [
          {
            "stage": "tool_selection",
            "context_characters": null,
            "context_character_limit": null,
            "context_bytes": 59586
          }
        ],
        "terminal_marker_exposures": 0,
        "reported_scripted_usage_complete": true
      },
      "correct": false,
      "whole_operation_seconds": 0.0018332920008106157,
      "returned_context_bytes": 403,
      "actual_model_usage": {
        "requests": 0,
        "models": {},
        "input_tokens": 0,
        "output_tokens": 0,
        "usage_complete": true
      },
      "arm": "baseline",
      "repeat": 2
    },
    {
      "case": "emoji_tool_exact",
      "runtime_version": "0.1.8",
      "runtime_loaded_from_requested_source": true,
      "fixture": {
        "character": "\ud83d\ude42",
        "target": "tool_selection",
        "characters": 15000,
        "marker_prefix_characters": 24,
        "payload_character_repetitions": 14862,
        "frozen_monotonic": 100.0,
        "max_steps": 2,
        "timeout_seconds": 120
      },
      "projected_context_sizes": {
        "context_projection": {
          "characters": 14964,
          "utf8_bytes": 59550
        },
        "tool_selection": {
          "characters": 15000,
          "utf8_bytes": 59586
        }
      },
      "expected": {
        "status": "done",
        "scripted_decisions": 1,
        "effects": 1,
        "intent_events": 1,
        "reason": null,
        "context_budget": null
      },
      "result": {
        "status": "done",
        "reason": null,
        "context_budget": null,
        "failure_metadata_correct": true,
        "success_metadata_correct": false,
        "behavior_correct": true,
        "scripted_decisions": 1,
        "effects": 1,
        "intent_events": 1,
        "native_inference_attempts": 0,
        "request_context_sizes": [
          {
            "stage": "tool_selection",
            "characters": 15000,
            "utf8_bytes": 59586
          }
        ],
        "returned_call_telemetry": [
          {
            "stage": "tool_selection",
            "context_characters": null,
            "context_character_limit": null,
            "context_bytes": 59586
          }
        ],
        "terminal_marker_exposures": 0,
        "reported_scripted_usage_complete": true
      },
      "correct": false,
      "whole_operation_seconds": 0.0020690419914899394,
      "returned_context_bytes": 403,
      "actual_model_usage": {
        "requests": 0,
        "models": {},
        "input_tokens": 0,
        "output_tokens": 0,
        "usage_complete": true
      },
      "arm": "baseline",
      "repeat": 3
    },
    {
      "case": "emoji_tool_exact",
      "runtime_version": "0.1.9",
      "runtime_loaded_from_requested_source": true,
      "fixture": {
        "character": "\ud83d\ude42",
        "target": "tool_selection",
        "characters": 15000,
        "marker_prefix_characters": 24,
        "payload_character_repetitions": 14862,
        "frozen_monotonic": 100.0,
        "max_steps": 2,
        "timeout_seconds": 120
      },
      "projected_context_sizes": {
        "context_projection": {
          "characters": 14964,
          "utf8_bytes": 59550
        },
        "tool_selection": {
          "characters": 15000,
          "utf8_bytes": 59586
        }
      },
      "expected": {
        "status": "done",
        "scripted_decisions": 1,
        "effects": 1,
        "intent_events": 1,
        "reason": null,
        "context_budget": null
      },
      "result": {
        "status": "done",
        "reason": null,
        "context_budget": null,
        "failure_metadata_correct": true,
        "success_metadata_correct": true,
        "behavior_correct": true,
        "scripted_decisions": 1,
        "effects": 1,
        "intent_events": 1,
        "native_inference_attempts": 0,
        "request_context_sizes": [
          {
            "stage": "tool_selection",
            "characters": 15000,
            "utf8_bytes": 59586
          }
        ],
        "returned_call_telemetry": [
          {
            "stage": "tool_selection",
            "context_characters": 15000,
            "context_character_limit": 15000,
            "context_bytes": 59586
          }
        ],
        "terminal_marker_exposures": 0,
        "reported_scripted_usage_complete": true
      },
      "correct": true,
      "whole_operation_seconds": 0.0019848340016324073,
      "returned_context_bytes": 466,
      "actual_model_usage": {
        "requests": 0,
        "models": {},
        "input_tokens": 0,
        "output_tokens": 0,
        "usage_complete": true
      },
      "arm": "candidate",
      "repeat": 3
    },
    {
      "case": "chinese_tool_over",
      "runtime_version": "0.1.8",
      "runtime_loaded_from_requested_source": true,
      "fixture": {
        "character": "\u4e2d",
        "target": "tool_selection",
        "characters": 15001,
        "marker_prefix_characters": 24,
        "payload_character_repetitions": 14863,
        "frozen_monotonic": 100.0,
        "max_steps": 2,
        "timeout_seconds": 120
      },
      "projected_context_sizes": {
        "context_projection": {
          "characters": 14965,
          "utf8_bytes": 44691
        },
        "tool_selection": {
          "characters": 15001,
          "utf8_bytes": 44727
        }
      },
      "expected": {
        "status": "needs_review",
        "scripted_decisions": 0,
        "effects": 0,
        "intent_events": 0,
        "reason": "context_budget_exceeded",
        "context_budget": {
          "scope": "loop_shared_context",
          "stage": "tool_selection",
          "unit": "json_characters",
          "limit": 15000,
          "observed": 15001
        }
      },
      "result": {
        "status": "needs_review",
        "reason": "context_budget_exceeded",
        "context_budget": null,
        "failure_metadata_correct": false,
        "success_metadata_correct": true,
        "behavior_correct": true,
        "scripted_decisions": 0,
        "effects": 0,
        "intent_events": 0,
        "native_inference_attempts": 0,
        "request_context_sizes": [],
        "returned_call_telemetry": [],
        "terminal_marker_exposures": 0,
        "reported_scripted_usage_complete": true
      },
      "correct": false,
      "whole_operation_seconds": 0.0015405409940285608,
      "returned_context_bytes": 259,
      "actual_model_usage": {
        "requests": 0,
        "models": {},
        "input_tokens": 0,
        "output_tokens": 0,
        "usage_complete": true
      },
      "arm": "baseline",
      "repeat": 1
    },
    {
      "case": "chinese_tool_over",
      "runtime_version": "0.1.9",
      "runtime_loaded_from_requested_source": true,
      "fixture": {
        "character": "\u4e2d",
        "target": "tool_selection",
        "characters": 15001,
        "marker_prefix_characters": 24,
        "payload_character_repetitions": 14863,
        "frozen_monotonic": 100.0,
        "max_steps": 2,
        "timeout_seconds": 120
      },
      "projected_context_sizes": {
        "context_projection": {
          "characters": 14965,
          "utf8_bytes": 44691
        },
        "tool_selection": {
          "characters": 15001,
          "utf8_bytes": 44727
        }
      },
      "expected": {
        "status": "needs_review",
        "scripted_decisions": 0,
        "effects": 0,
        "intent_events": 0,
        "reason": "context_budget_exceeded",
        "context_budget": {
          "scope": "loop_shared_context",
          "stage": "tool_selection",
          "unit": "json_characters",
          "limit": 15000,
          "observed": 15001
        }
      },
      "result": {
        "status": "needs_review",
        "reason": "context_budget_exceeded",
        "context_budget": {
          "scope": "loop_shared_context",
          "stage": "tool_selection",
          "unit": "json_characters",
          "limit": 15000,
          "observed": 15001
        },
        "failure_metadata_correct": true,
        "success_metadata_correct": true,
        "behavior_correct": true,
        "scripted_decisions": 0,
        "effects": 0,
        "intent_events": 0,
        "native_inference_attempts": 0,
        "request_context_sizes": [],
        "returned_call_telemetry": [],
        "terminal_marker_exposures": 0,
        "reported_scripted_usage_complete": true
      },
      "correct": true,
      "whole_operation_seconds": 0.001431666998541914,
      "returned_context_bytes": 400,
      "actual_model_usage": {
        "requests": 0,
        "models": {},
        "input_tokens": 0,
        "output_tokens": 0,
        "usage_complete": true
      },
      "arm": "candidate",
      "repeat": 1
    },
    {
      "case": "chinese_tool_over",
      "runtime_version": "0.1.9",
      "runtime_loaded_from_requested_source": true,
      "fixture": {
        "character": "\u4e2d",
        "target": "tool_selection",
        "characters": 15001,
        "marker_prefix_characters": 24,
        "payload_character_repetitions": 14863,
        "frozen_monotonic": 100.0,
        "max_steps": 2,
        "timeout_seconds": 120
      },
      "projected_context_sizes": {
        "context_projection": {
          "characters": 14965,
          "utf8_bytes": 44691
        },
        "tool_selection": {
          "characters": 15001,
          "utf8_bytes": 44727
        }
      },
      "expected": {
        "status": "needs_review",
        "scripted_decisions": 0,
        "effects": 0,
        "intent_events": 0,
        "reason": "context_budget_exceeded",
        "context_budget": {
          "scope": "loop_shared_context",
          "stage": "tool_selection",
          "unit": "json_characters",
          "limit": 15000,
          "observed": 15001
        }
      },
      "result": {
        "status": "needs_review",
        "reason": "context_budget_exceeded",
        "context_budget": {
          "scope": "loop_shared_context",
          "stage": "tool_selection",
          "unit": "json_characters",
          "limit": 15000,
          "observed": 15001
        },
        "failure_metadata_correct": true,
        "success_metadata_correct": true,
        "behavior_correct": true,
        "scripted_decisions": 0,
        "effects": 0,
        "intent_events": 0,
        "native_inference_attempts": 0,
        "request_context_sizes": [],
        "returned_call_telemetry": [],
        "terminal_marker_exposures": 0,
        "reported_scripted_usage_complete": true
      },
      "correct": true,
      "whole_operation_seconds": 0.0015132920088944957,
      "returned_context_bytes": 400,
      "actual_model_usage": {
        "requests": 0,
        "models": {},
        "input_tokens": 0,
        "output_tokens": 0,
        "usage_complete": true
      },
      "arm": "candidate",
      "repeat": 2
    },
    {
      "case": "chinese_tool_over",
      "runtime_version": "0.1.8",
      "runtime_loaded_from_requested_source": true,
      "fixture": {
        "character": "\u4e2d",
        "target": "tool_selection",
        "characters": 15001,
        "marker_prefix_characters": 24,
        "payload_character_repetitions": 14863,
        "frozen_monotonic": 100.0,
        "max_steps": 2,
        "timeout_seconds": 120
      },
      "projected_context_sizes": {
        "context_projection": {
          "characters": 14965,
          "utf8_bytes": 44691
        },
        "tool_selection": {
          "characters": 15001,
          "utf8_bytes": 44727
        }
      },
      "expected": {
        "status": "needs_review",
        "scripted_decisions": 0,
        "effects": 0,
        "intent_events": 0,
        "reason": "context_budget_exceeded",
        "context_budget": {
          "scope": "loop_shared_context",
          "stage": "tool_selection",
          "unit": "json_characters",
          "limit": 15000,
          "observed": 15001
        }
      },
      "result": {
        "status": "needs_review",
        "reason": "context_budget_exceeded",
        "context_budget": null,
        "failure_metadata_correct": false,
        "success_metadata_correct": true,
        "behavior_correct": true,
        "scripted_decisions": 0,
        "effects": 0,
        "intent_events": 0,
        "native_inference_attempts": 0,
        "request_context_sizes": [],
        "returned_call_telemetry": [],
        "terminal_marker_exposures": 0,
        "reported_scripted_usage_complete": true
      },
      "correct": false,
      "whole_operation_seconds": 0.0014635420084232464,
      "returned_context_bytes": 259,
      "actual_model_usage": {
        "requests": 0,
        "models": {},
        "input_tokens": 0,
        "output_tokens": 0,
        "usage_complete": true
      },
      "arm": "baseline",
      "repeat": 2
    },
    {
      "case": "chinese_tool_over",
      "runtime_version": "0.1.8",
      "runtime_loaded_from_requested_source": true,
      "fixture": {
        "character": "\u4e2d",
        "target": "tool_selection",
        "characters": 15001,
        "marker_prefix_characters": 24,
        "payload_character_repetitions": 14863,
        "frozen_monotonic": 100.0,
        "max_steps": 2,
        "timeout_seconds": 120
      },
      "projected_context_sizes": {
        "context_projection": {
          "characters": 14965,
          "utf8_bytes": 44691
        },
        "tool_selection": {
          "characters": 15001,
          "utf8_bytes": 44727
        }
      },
      "expected": {
        "status": "needs_review",
        "scripted_decisions": 0,
        "effects": 0,
        "intent_events": 0,
        "reason": "context_budget_exceeded",
        "context_budget": {
          "scope": "loop_shared_context",
          "stage": "tool_selection",
          "unit": "json_characters",
          "limit": 15000,
          "observed": 15001
        }
      },
      "result": {
        "status": "needs_review",
        "reason": "context_budget_exceeded",
        "context_budget": null,
        "failure_metadata_correct": false,
        "success_metadata_correct": true,
        "behavior_correct": true,
        "scripted_decisions": 0,
        "effects": 0,
        "intent_events": 0,
        "native_inference_attempts": 0,
        "request_context_sizes": [],
        "returned_call_telemetry": [],
        "terminal_marker_exposures": 0,
        "reported_scripted_usage_complete": true
      },
      "correct": false,
      "whole_operation_seconds": 0.0013531250006053597,
      "returned_context_bytes": 259,
      "actual_model_usage": {
        "requests": 0,
        "models": {},
        "input_tokens": 0,
        "output_tokens": 0,
        "usage_complete": true
      },
      "arm": "baseline",
      "repeat": 3
    },
    {
      "case": "chinese_tool_over",
      "runtime_version": "0.1.9",
      "runtime_loaded_from_requested_source": true,
      "fixture": {
        "character": "\u4e2d",
        "target": "tool_selection",
        "characters": 15001,
        "marker_prefix_characters": 24,
        "payload_character_repetitions": 14863,
        "frozen_monotonic": 100.0,
        "max_steps": 2,
        "timeout_seconds": 120
      },
      "projected_context_sizes": {
        "context_projection": {
          "characters": 14965,
          "utf8_bytes": 44691
        },
        "tool_selection": {
          "characters": 15001,
          "utf8_bytes": 44727
        }
      },
      "expected": {
        "status": "needs_review",
        "scripted_decisions": 0,
        "effects": 0,
        "intent_events": 0,
        "reason": "context_budget_exceeded",
        "context_budget": {
          "scope": "loop_shared_context",
          "stage": "tool_selection",
          "unit": "json_characters",
          "limit": 15000,
          "observed": 15001
        }
      },
      "result": {
        "status": "needs_review",
        "reason": "context_budget_exceeded",
        "context_budget": {
          "scope": "loop_shared_context",
          "stage": "tool_selection",
          "unit": "json_characters",
          "limit": 15000,
          "observed": 15001
        },
        "failure_metadata_correct": true,
        "success_metadata_correct": true,
        "behavior_correct": true,
        "scripted_decisions": 0,
        "effects": 0,
        "intent_events": 0,
        "native_inference_attempts": 0,
        "request_context_sizes": [],
        "returned_call_telemetry": [],
        "terminal_marker_exposures": 0,
        "reported_scripted_usage_complete": true
      },
      "correct": true,
      "whole_operation_seconds": 0.0011881250102305785,
      "returned_context_bytes": 400,
      "actual_model_usage": {
        "requests": 0,
        "models": {},
        "input_tokens": 0,
        "output_tokens": 0,
        "usage_complete": true
      },
      "arm": "candidate",
      "repeat": 3
    },
    {
      "case": "emoji_projection_over",
      "runtime_version": "0.1.8",
      "runtime_loaded_from_requested_source": true,
      "fixture": {
        "character": "\ud83d\ude42",
        "target": "context_projection",
        "characters": 15001,
        "marker_prefix_characters": 24,
        "payload_character_repetitions": 14899,
        "frozen_monotonic": 100.0,
        "max_steps": 2,
        "timeout_seconds": 120
      },
      "projected_context_sizes": {
        "context_projection": {
          "characters": 15001,
          "utf8_bytes": 59698
        },
        "tool_selection": {
          "characters": 15037,
          "utf8_bytes": 59734
        }
      },
      "expected": {
        "status": "needs_review",
        "scripted_decisions": 0,
        "effects": 0,
        "intent_events": 0,
        "reason": "context_budget_exceeded",
        "context_budget": {
          "scope": "loop_shared_context",
          "stage": "context_projection",
          "unit": "json_characters",
          "limit": 15000,
          "observed": 15001
        }
      },
      "result": {
        "status": "needs_review",
        "reason": "context_budget_exceeded",
        "context_budget": null,
        "failure_metadata_correct": false,
        "success_metadata_correct": true,
        "behavior_correct": true,
        "scripted_decisions": 0,
        "effects": 0,
        "intent_events": 0,
        "native_inference_attempts": 0,
        "request_context_sizes": [],
        "returned_call_telemetry": [],
        "terminal_marker_exposures": 0,
        "reported_scripted_usage_complete": true
      },
      "correct": false,
      "whole_operation_seconds": 0.001269207990844734,
      "returned_context_bytes": 259,
      "actual_model_usage": {
        "requests": 0,
        "models": {},
        "input_tokens": 0,
        "output_tokens": 0,
        "usage_complete": true
      },
      "arm": "baseline",
      "repeat": 1
    },
    {
      "case": "emoji_projection_over",
      "runtime_version": "0.1.9",
      "runtime_loaded_from_requested_source": true,
      "fixture": {
        "character": "\ud83d\ude42",
        "target": "context_projection",
        "characters": 15001,
        "marker_prefix_characters": 24,
        "payload_character_repetitions": 14899,
        "frozen_monotonic": 100.0,
        "max_steps": 2,
        "timeout_seconds": 120
      },
      "projected_context_sizes": {
        "context_projection": {
          "characters": 15001,
          "utf8_bytes": 59698
        },
        "tool_selection": {
          "characters": 15037,
          "utf8_bytes": 59734
        }
      },
      "expected": {
        "status": "needs_review",
        "scripted_decisions": 0,
        "effects": 0,
        "intent_events": 0,
        "reason": "context_budget_exceeded",
        "context_budget": {
          "scope": "loop_shared_context",
          "stage": "context_projection",
          "unit": "json_characters",
          "limit": 15000,
          "observed": 15001
        }
      },
      "result": {
        "status": "needs_review",
        "reason": "context_budget_exceeded",
        "context_budget": {
          "scope": "loop_shared_context",
          "stage": "context_projection",
          "unit": "json_characters",
          "limit": 15000,
          "observed": 15001
        },
        "failure_metadata_correct": true,
        "success_metadata_correct": true,
        "behavior_correct": true,
        "scripted_decisions": 0,
        "effects": 0,
        "intent_events": 0,
        "native_inference_attempts": 0,
        "request_context_sizes": [],
        "returned_call_telemetry": [],
        "terminal_marker_exposures": 0,
        "reported_scripted_usage_complete": true
      },
      "correct": true,
      "whole_operation_seconds": 0.0012994579883525148,
      "returned_context_bytes": 404,
      "actual_model_usage": {
        "requests": 0,
        "models": {},
        "input_tokens": 0,
        "output_tokens": 0,
        "usage_complete": true
      },
      "arm": "candidate",
      "repeat": 1
    },
    {
      "case": "emoji_projection_over",
      "runtime_version": "0.1.9",
      "runtime_loaded_from_requested_source": true,
      "fixture": {
        "character": "\ud83d\ude42",
        "target": "context_projection",
        "characters": 15001,
        "marker_prefix_characters": 24,
        "payload_character_repetitions": 14899,
        "frozen_monotonic": 100.0,
        "max_steps": 2,
        "timeout_seconds": 120
      },
      "projected_context_sizes": {
        "context_projection": {
          "characters": 15001,
          "utf8_bytes": 59698
        },
        "tool_selection": {
          "characters": 15037,
          "utf8_bytes": 59734
        }
      },
      "expected": {
        "status": "needs_review",
        "scripted_decisions": 0,
        "effects": 0,
        "intent_events": 0,
        "reason": "context_budget_exceeded",
        "context_budget": {
          "scope": "loop_shared_context",
          "stage": "context_projection",
          "unit": "json_characters",
          "limit": 15000,
          "observed": 15001
        }
      },
      "result": {
        "status": "needs_review",
        "reason": "context_budget_exceeded",
        "context_budget": {
          "scope": "loop_shared_context",
          "stage": "context_projection",
          "unit": "json_characters",
          "limit": 15000,
          "observed": 15001
        },
        "failure_metadata_correct": true,
        "success_metadata_correct": true,
        "behavior_correct": true,
        "scripted_decisions": 0,
        "effects": 0,
        "intent_events": 0,
        "native_inference_attempts": 0,
        "request_context_sizes": [],
        "returned_call_telemetry": [],
        "terminal_marker_exposures": 0,
        "reported_scripted_usage_complete": true
      },
      "correct": true,
      "whole_operation_seconds": 0.0012457499979063869,
      "returned_context_bytes": 404,
      "actual_model_usage": {
        "requests": 0,
        "models": {},
        "input_tokens": 0,
        "output_tokens": 0,
        "usage_complete": true
      },
      "arm": "candidate",
      "repeat": 2
    },
    {
      "case": "emoji_projection_over",
      "runtime_version": "0.1.8",
      "runtime_loaded_from_requested_source": true,
      "fixture": {
        "character": "\ud83d\ude42",
        "target": "context_projection",
        "characters": 15001,
        "marker_prefix_characters": 24,
        "payload_character_repetitions": 14899,
        "frozen_monotonic": 100.0,
        "max_steps": 2,
        "timeout_seconds": 120
      },
      "projected_context_sizes": {
        "context_projection": {
          "characters": 15001,
          "utf8_bytes": 59698
        },
        "tool_selection": {
          "characters": 15037,
          "utf8_bytes": 59734
        }
      },
      "expected": {
        "status": "needs_review",
        "scripted_decisions": 0,
        "effects": 0,
        "intent_events": 0,
        "reason": "context_budget_exceeded",
        "context_budget": {
          "scope": "loop_shared_context",
          "stage": "context_projection",
          "unit": "json_characters",
          "limit": 15000,
          "observed": 15001
        }
      },
      "result": {
        "status": "needs_review",
        "reason": "context_budget_exceeded",
        "context_budget": null,
        "failure_metadata_correct": false,
        "success_metadata_correct": true,
        "behavior_correct": true,
        "scripted_decisions": 0,
        "effects": 0,
        "intent_events": 0,
        "native_inference_attempts": 0,
        "request_context_sizes": [],
        "returned_call_telemetry": [],
        "terminal_marker_exposures": 0,
        "reported_scripted_usage_complete": true
      },
      "correct": false,
      "whole_operation_seconds": 0.001229041998158209,
      "returned_context_bytes": 259,
      "actual_model_usage": {
        "requests": 0,
        "models": {},
        "input_tokens": 0,
        "output_tokens": 0,
        "usage_complete": true
      },
      "arm": "baseline",
      "repeat": 2
    },
    {
      "case": "emoji_projection_over",
      "runtime_version": "0.1.8",
      "runtime_loaded_from_requested_source": true,
      "fixture": {
        "character": "\ud83d\ude42",
        "target": "context_projection",
        "characters": 15001,
        "marker_prefix_characters": 24,
        "payload_character_repetitions": 14899,
        "frozen_monotonic": 100.0,
        "max_steps": 2,
        "timeout_seconds": 120
      },
      "projected_context_sizes": {
        "context_projection": {
          "characters": 15001,
          "utf8_bytes": 59698
        },
        "tool_selection": {
          "characters": 15037,
          "utf8_bytes": 59734
        }
      },
      "expected": {
        "status": "needs_review",
        "scripted_decisions": 0,
        "effects": 0,
        "intent_events": 0,
        "reason": "context_budget_exceeded",
        "context_budget": {
          "scope": "loop_shared_context",
          "stage": "context_projection",
          "unit": "json_characters",
          "limit": 15000,
          "observed": 15001
        }
      },
      "result": {
        "status": "needs_review",
        "reason": "context_budget_exceeded",
        "context_budget": null,
        "failure_metadata_correct": false,
        "success_metadata_correct": true,
        "behavior_correct": true,
        "scripted_decisions": 0,
        "effects": 0,
        "intent_events": 0,
        "native_inference_attempts": 0,
        "request_context_sizes": [],
        "returned_call_telemetry": [],
        "terminal_marker_exposures": 0,
        "reported_scripted_usage_complete": true
      },
      "correct": false,
      "whole_operation_seconds": 0.0012246670085005462,
      "returned_context_bytes": 259,
      "actual_model_usage": {
        "requests": 0,
        "models": {},
        "input_tokens": 0,
        "output_tokens": 0,
        "usage_complete": true
      },
      "arm": "baseline",
      "repeat": 3
    },
    {
      "case": "emoji_projection_over",
      "runtime_version": "0.1.9",
      "runtime_loaded_from_requested_source": true,
      "fixture": {
        "character": "\ud83d\ude42",
        "target": "context_projection",
        "characters": 15001,
        "marker_prefix_characters": 24,
        "payload_character_repetitions": 14899,
        "frozen_monotonic": 100.0,
        "max_steps": 2,
        "timeout_seconds": 120
      },
      "projected_context_sizes": {
        "context_projection": {
          "characters": 15001,
          "utf8_bytes": 59698
        },
        "tool_selection": {
          "characters": 15037,
          "utf8_bytes": 59734
        }
      },
      "expected": {
        "status": "needs_review",
        "scripted_decisions": 0,
        "effects": 0,
        "intent_events": 0,
        "reason": "context_budget_exceeded",
        "context_budget": {
          "scope": "loop_shared_context",
          "stage": "context_projection",
          "unit": "json_characters",
          "limit": 15000,
          "observed": 15001
        }
      },
      "result": {
        "status": "needs_review",
        "reason": "context_budget_exceeded",
        "context_budget": {
          "scope": "loop_shared_context",
          "stage": "context_projection",
          "unit": "json_characters",
          "limit": 15000,
          "observed": 15001
        },
        "failure_metadata_correct": true,
        "success_metadata_correct": true,
        "behavior_correct": true,
        "scripted_decisions": 0,
        "effects": 0,
        "intent_events": 0,
        "native_inference_attempts": 0,
        "request_context_sizes": [],
        "returned_call_telemetry": [],
        "terminal_marker_exposures": 0,
        "reported_scripted_usage_complete": true
      },
      "correct": true,
      "whole_operation_seconds": 0.0011468750017229468,
      "returned_context_bytes": 404,
      "actual_model_usage": {
        "requests": 0,
        "models": {},
        "input_tokens": 0,
        "output_tokens": 0,
        "usage_complete": true
      },
      "arm": "candidate",
      "repeat": 3
    },
    {
      "case": "ascii_parameter_exact",
      "runtime_version": "0.1.8",
      "runtime_loaded_from_requested_source": true,
      "fixture": {
        "character": "A",
        "target": "parameter_selection",
        "characters": 15000,
        "marker_prefix_characters": 24,
        "payload_character_repetitions": 14537,
        "frozen_monotonic": 100.0,
        "max_steps": 2,
        "timeout_seconds": 120
      },
      "projected_context_sizes": {
        "context_projection": {
          "characters": 14639,
          "utf8_bytes": 14639
        },
        "tool_selection": {
          "characters": 14675,
          "utf8_bytes": 14675
        },
        "parameter_selection": {
          "characters": 15000,
          "utf8_bytes": 15000
        }
      },
      "expected": {
        "status": "done",
        "scripted_decisions": 2,
        "effects": 1,
        "intent_events": 1,
        "reason": null,
        "context_budget": null
      },
      "result": {
        "status": "done",
        "reason": null,
        "context_budget": null,
        "failure_metadata_correct": true,
        "success_metadata_correct": false,
        "behavior_correct": true,
        "scripted_decisions": 2,
        "effects": 1,
        "intent_events": 1,
        "native_inference_attempts": 0,
        "request_context_sizes": [
          {
            "stage": "tool_selection",
            "characters": 14675,
            "utf8_bytes": 14675
          },
          {
            "stage": "parameter_selection",
            "characters": 15000,
            "utf8_bytes": 15000
          }
        ],
        "returned_call_telemetry": [
          {
            "stage": "tool_selection",
            "context_characters": null,
            "context_character_limit": null,
            "context_bytes": 14675
          },
          {
            "stage": "parameter_selection",
            "context_characters": null,
            "context_character_limit": null,
            "context_bytes": 15000
          }
        ],
        "terminal_marker_exposures": 0,
        "reported_scripted_usage_complete": true
      },
      "correct": false,
      "whole_operation_seconds": 0.002291915996465832,
      "returned_context_bytes": 645,
      "actual_model_usage": {
        "requests": 0,
        "models": {},
        "input_tokens": 0,
        "output_tokens": 0,
        "usage_complete": true
      },
      "arm": "baseline",
      "repeat": 1
    },
    {
      "case": "ascii_parameter_exact",
      "runtime_version": "0.1.9",
      "runtime_loaded_from_requested_source": true,
      "fixture": {
        "character": "A",
        "target": "parameter_selection",
        "characters": 15000,
        "marker_prefix_characters": 24,
        "payload_character_repetitions": 14537,
        "frozen_monotonic": 100.0,
        "max_steps": 2,
        "timeout_seconds": 120
      },
      "projected_context_sizes": {
        "context_projection": {
          "characters": 14639,
          "utf8_bytes": 14639
        },
        "tool_selection": {
          "characters": 14675,
          "utf8_bytes": 14675
        },
        "parameter_selection": {
          "characters": 15000,
          "utf8_bytes": 15000
        }
      },
      "expected": {
        "status": "done",
        "scripted_decisions": 2,
        "effects": 1,
        "intent_events": 1,
        "reason": null,
        "context_budget": null
      },
      "result": {
        "status": "done",
        "reason": null,
        "context_budget": null,
        "failure_metadata_correct": true,
        "success_metadata_correct": true,
        "behavior_correct": true,
        "scripted_decisions": 2,
        "effects": 1,
        "intent_events": 1,
        "native_inference_attempts": 0,
        "request_context_sizes": [
          {
            "stage": "tool_selection",
            "characters": 14675,
            "utf8_bytes": 14675
          },
          {
            "stage": "parameter_selection",
            "characters": 15000,
            "utf8_bytes": 15000
          }
        ],
        "returned_call_telemetry": [
          {
            "stage": "tool_selection",
            "context_characters": 14675,
            "context_character_limit": 15000,
            "context_bytes": 14675
          },
          {
            "stage": "parameter_selection",
            "context_characters": 15000,
            "context_character_limit": 15000,
            "context_bytes": 15000
          }
        ],
        "terminal_marker_exposures": 0,
        "reported_scripted_usage_complete": true
      },
      "correct": true,
      "whole_operation_seconds": 0.0020206250046612695,
      "returned_context_bytes": 771,
      "actual_model_usage": {
        "requests": 0,
        "models": {},
        "input_tokens": 0,
        "output_tokens": 0,
        "usage_complete": true
      },
      "arm": "candidate",
      "repeat": 1
    },
    {
      "case": "ascii_parameter_exact",
      "runtime_version": "0.1.9",
      "runtime_loaded_from_requested_source": true,
      "fixture": {
        "character": "A",
        "target": "parameter_selection",
        "characters": 15000,
        "marker_prefix_characters": 24,
        "payload_character_repetitions": 14537,
        "frozen_monotonic": 100.0,
        "max_steps": 2,
        "timeout_seconds": 120
      },
      "projected_context_sizes": {
        "context_projection": {
          "characters": 14639,
          "utf8_bytes": 14639
        },
        "tool_selection": {
          "characters": 14675,
          "utf8_bytes": 14675
        },
        "parameter_selection": {
          "characters": 15000,
          "utf8_bytes": 15000
        }
      },
      "expected": {
        "status": "done",
        "scripted_decisions": 2,
        "effects": 1,
        "intent_events": 1,
        "reason": null,
        "context_budget": null
      },
      "result": {
        "status": "done",
        "reason": null,
        "context_budget": null,
        "failure_metadata_correct": true,
        "success_metadata_correct": true,
        "behavior_correct": true,
        "scripted_decisions": 2,
        "effects": 1,
        "intent_events": 1,
        "native_inference_attempts": 0,
        "request_context_sizes": [
          {
            "stage": "tool_selection",
            "characters": 14675,
            "utf8_bytes": 14675
          },
          {
            "stage": "parameter_selection",
            "characters": 15000,
            "utf8_bytes": 15000
          }
        ],
        "returned_call_telemetry": [
          {
            "stage": "tool_selection",
            "context_characters": 14675,
            "context_character_limit": 15000,
            "context_bytes": 14675
          },
          {
            "stage": "parameter_selection",
            "context_characters": 15000,
            "context_character_limit": 15000,
            "context_bytes": 15000
          }
        ],
        "terminal_marker_exposures": 0,
        "reported_scripted_usage_complete": true
      },
      "correct": true,
      "whole_operation_seconds": 0.002188917002058588,
      "returned_context_bytes": 771,
      "actual_model_usage": {
        "requests": 0,
        "models": {},
        "input_tokens": 0,
        "output_tokens": 0,
        "usage_complete": true
      },
      "arm": "candidate",
      "repeat": 2
    },
    {
      "case": "ascii_parameter_exact",
      "runtime_version": "0.1.8",
      "runtime_loaded_from_requested_source": true,
      "fixture": {
        "character": "A",
        "target": "parameter_selection",
        "characters": 15000,
        "marker_prefix_characters": 24,
        "payload_character_repetitions": 14537,
        "frozen_monotonic": 100.0,
        "max_steps": 2,
        "timeout_seconds": 120
      },
      "projected_context_sizes": {
        "context_projection": {
          "characters": 14639,
          "utf8_bytes": 14639
        },
        "tool_selection": {
          "characters": 14675,
          "utf8_bytes": 14675
        },
        "parameter_selection": {
          "characters": 15000,
          "utf8_bytes": 15000
        }
      },
      "expected": {
        "status": "done",
        "scripted_decisions": 2,
        "effects": 1,
        "intent_events": 1,
        "reason": null,
        "context_budget": null
      },
      "result": {
        "status": "done",
        "reason": null,
        "context_budget": null,
        "failure_metadata_correct": true,
        "success_metadata_correct": false,
        "behavior_correct": true,
        "scripted_decisions": 2,
        "effects": 1,
        "intent_events": 1,
        "native_inference_attempts": 0,
        "request_context_sizes": [
          {
            "stage": "tool_selection",
            "characters": 14675,
            "utf8_bytes": 14675
          },
          {
            "stage": "parameter_selection",
            "characters": 15000,
            "utf8_bytes": 15000
          }
        ],
        "returned_call_telemetry": [
          {
            "stage": "tool_selection",
            "context_characters": null,
            "context_character_limit": null,
            "context_bytes": 14675
          },
          {
            "stage": "parameter_selection",
            "context_characters": null,
            "context_character_limit": null,
            "context_bytes": 15000
          }
        ],
        "terminal_marker_exposures": 0,
        "reported_scripted_usage_complete": true
      },
      "correct": false,
      "whole_operation_seconds": 0.002442208002321422,
      "returned_context_bytes": 645,
      "actual_model_usage": {
        "requests": 0,
        "models": {},
        "input_tokens": 0,
        "output_tokens": 0,
        "usage_complete": true
      },
      "arm": "baseline",
      "repeat": 2
    },
    {
      "case": "ascii_parameter_exact",
      "runtime_version": "0.1.8",
      "runtime_loaded_from_requested_source": true,
      "fixture": {
        "character": "A",
        "target": "parameter_selection",
        "characters": 15000,
        "marker_prefix_characters": 24,
        "payload_character_repetitions": 14537,
        "frozen_monotonic": 100.0,
        "max_steps": 2,
        "timeout_seconds": 120
      },
      "projected_context_sizes": {
        "context_projection": {
          "characters": 14639,
          "utf8_bytes": 14639
        },
        "tool_selection": {
          "characters": 14675,
          "utf8_bytes": 14675
        },
        "parameter_selection": {
          "characters": 15000,
          "utf8_bytes": 15000
        }
      },
      "expected": {
        "status": "done",
        "scripted_decisions": 2,
        "effects": 1,
        "intent_events": 1,
        "reason": null,
        "context_budget": null
      },
      "result": {
        "status": "done",
        "reason": null,
        "context_budget": null,
        "failure_metadata_correct": true,
        "success_metadata_correct": false,
        "behavior_correct": true,
        "scripted_decisions": 2,
        "effects": 1,
        "intent_events": 1,
        "native_inference_attempts": 0,
        "request_context_sizes": [
          {
            "stage": "tool_selection",
            "characters": 14675,
            "utf8_bytes": 14675
          },
          {
            "stage": "parameter_selection",
            "characters": 15000,
            "utf8_bytes": 15000
          }
        ],
        "returned_call_telemetry": [
          {
            "stage": "tool_selection",
            "context_characters": null,
            "context_character_limit": null,
            "context_bytes": 14675
          },
          {
            "stage": "parameter_selection",
            "context_characters": null,
            "context_character_limit": null,
            "context_bytes": 15000
          }
        ],
        "terminal_marker_exposures": 0,
        "reported_scripted_usage_complete": true
      },
      "correct": false,
      "whole_operation_seconds": 0.002171375002944842,
      "returned_context_bytes": 645,
      "actual_model_usage": {
        "requests": 0,
        "models": {},
        "input_tokens": 0,
        "output_tokens": 0,
        "usage_complete": true
      },
      "arm": "baseline",
      "repeat": 3
    },
    {
      "case": "ascii_parameter_exact",
      "runtime_version": "0.1.9",
      "runtime_loaded_from_requested_source": true,
      "fixture": {
        "character": "A",
        "target": "parameter_selection",
        "characters": 15000,
        "marker_prefix_characters": 24,
        "payload_character_repetitions": 14537,
        "frozen_monotonic": 100.0,
        "max_steps": 2,
        "timeout_seconds": 120
      },
      "projected_context_sizes": {
        "context_projection": {
          "characters": 14639,
          "utf8_bytes": 14639
        },
        "tool_selection": {
          "characters": 14675,
          "utf8_bytes": 14675
        },
        "parameter_selection": {
          "characters": 15000,
          "utf8_bytes": 15000
        }
      },
      "expected": {
        "status": "done",
        "scripted_decisions": 2,
        "effects": 1,
        "intent_events": 1,
        "reason": null,
        "context_budget": null
      },
      "result": {
        "status": "done",
        "reason": null,
        "context_budget": null,
        "failure_metadata_correct": true,
        "success_metadata_correct": true,
        "behavior_correct": true,
        "scripted_decisions": 2,
        "effects": 1,
        "intent_events": 1,
        "native_inference_attempts": 0,
        "request_context_sizes": [
          {
            "stage": "tool_selection",
            "characters": 14675,
            "utf8_bytes": 14675
          },
          {
            "stage": "parameter_selection",
            "characters": 15000,
            "utf8_bytes": 15000
          }
        ],
        "returned_call_telemetry": [
          {
            "stage": "tool_selection",
            "context_characters": 14675,
            "context_character_limit": 15000,
            "context_bytes": 14675
          },
          {
            "stage": "parameter_selection",
            "context_characters": 15000,
            "context_character_limit": 15000,
            "context_bytes": 15000
          }
        ],
        "terminal_marker_exposures": 0,
        "reported_scripted_usage_complete": true
      },
      "correct": true,
      "whole_operation_seconds": 0.002236749991425313,
      "returned_context_bytes": 771,
      "actual_model_usage": {
        "requests": 0,
        "models": {},
        "input_tokens": 0,
        "output_tokens": 0,
        "usage_complete": true
      },
      "arm": "candidate",
      "repeat": 3
    },
    {
      "case": "ascii_parameter_over",
      "runtime_version": "0.1.8",
      "runtime_loaded_from_requested_source": true,
      "fixture": {
        "character": "A",
        "target": "parameter_selection",
        "characters": 15001,
        "marker_prefix_characters": 24,
        "payload_character_repetitions": 14538,
        "frozen_monotonic": 100.0,
        "max_steps": 2,
        "timeout_seconds": 120
      },
      "projected_context_sizes": {
        "context_projection": {
          "characters": 14640,
          "utf8_bytes": 14640
        },
        "tool_selection": {
          "characters": 14676,
          "utf8_bytes": 14676
        },
        "parameter_selection": {
          "characters": 15001,
          "utf8_bytes": 15001
        }
      },
      "expected": {
        "status": "needs_review",
        "scripted_decisions": 1,
        "effects": 0,
        "intent_events": 0,
        "reason": "context_budget_exceeded",
        "context_budget": {
          "scope": "loop_shared_context",
          "stage": "parameter_selection",
          "unit": "json_characters",
          "limit": 15000,
          "observed": 15001
        }
      },
      "result": {
        "status": "needs_review",
        "reason": "context_budget_exceeded",
        "context_budget": null,
        "failure_metadata_correct": false,
        "success_metadata_correct": false,
        "behavior_correct": true,
        "scripted_decisions": 1,
        "effects": 0,
        "intent_events": 0,
        "native_inference_attempts": 0,
        "request_context_sizes": [
          {
            "stage": "tool_selection",
            "characters": 14676,
            "utf8_bytes": 14676
          }
        ],
        "returned_call_telemetry": [
          {
            "stage": "tool_selection",
            "context_characters": null,
            "context_character_limit": null,
            "context_bytes": 14676
          }
        ],
        "terminal_marker_exposures": 0,
        "reported_scripted_usage_complete": true
      },
      "correct": false,
      "whole_operation_seconds": 0.001727958006085828,
      "returned_context_bytes": 449,
      "actual_model_usage": {
        "requests": 0,
        "models": {},
        "input_tokens": 0,
        "output_tokens": 0,
        "usage_complete": true
      },
      "arm": "baseline",
      "repeat": 1
    },
    {
      "case": "ascii_parameter_over",
      "runtime_version": "0.1.9",
      "runtime_loaded_from_requested_source": true,
      "fixture": {
        "character": "A",
        "target": "parameter_selection",
        "characters": 15001,
        "marker_prefix_characters": 24,
        "payload_character_repetitions": 14538,
        "frozen_monotonic": 100.0,
        "max_steps": 2,
        "timeout_seconds": 120
      },
      "projected_context_sizes": {
        "context_projection": {
          "characters": 14640,
          "utf8_bytes": 14640
        },
        "tool_selection": {
          "characters": 14676,
          "utf8_bytes": 14676
        },
        "parameter_selection": {
          "characters": 15001,
          "utf8_bytes": 15001
        }
      },
      "expected": {
        "status": "needs_review",
        "scripted_decisions": 1,
        "effects": 0,
        "intent_events": 0,
        "reason": "context_budget_exceeded",
        "context_budget": {
          "scope": "loop_shared_context",
          "stage": "parameter_selection",
          "unit": "json_characters",
          "limit": 15000,
          "observed": 15001
        }
      },
      "result": {
        "status": "needs_review",
        "reason": "context_budget_exceeded",
        "context_budget": {
          "scope": "loop_shared_context",
          "stage": "parameter_selection",
          "unit": "json_characters",
          "limit": 15000,
          "observed": 15001
        },
        "failure_metadata_correct": true,
        "success_metadata_correct": true,
        "behavior_correct": true,
        "scripted_decisions": 1,
        "effects": 0,
        "intent_events": 0,
        "native_inference_attempts": 0,
        "request_context_sizes": [
          {
            "stage": "tool_selection",
            "characters": 14676,
            "utf8_bytes": 14676
          }
        ],
        "returned_call_telemetry": [
          {
            "stage": "tool_selection",
            "context_characters": 14676,
            "context_character_limit": 15000,
            "context_bytes": 14676
          }
        ],
        "terminal_marker_exposures": 0,
        "reported_scripted_usage_complete": true
      },
      "correct": true,
      "whole_operation_seconds": 0.002262000009068288,
      "returned_context_bytes": 658,
      "actual_model_usage": {
        "requests": 0,
        "models": {},
        "input_tokens": 0,
        "output_tokens": 0,
        "usage_complete": true
      },
      "arm": "candidate",
      "repeat": 1
    },
    {
      "case": "ascii_parameter_over",
      "runtime_version": "0.1.9",
      "runtime_loaded_from_requested_source": true,
      "fixture": {
        "character": "A",
        "target": "parameter_selection",
        "characters": 15001,
        "marker_prefix_characters": 24,
        "payload_character_repetitions": 14538,
        "frozen_monotonic": 100.0,
        "max_steps": 2,
        "timeout_seconds": 120
      },
      "projected_context_sizes": {
        "context_projection": {
          "characters": 14640,
          "utf8_bytes": 14640
        },
        "tool_selection": {
          "characters": 14676,
          "utf8_bytes": 14676
        },
        "parameter_selection": {
          "characters": 15001,
          "utf8_bytes": 15001
        }
      },
      "expected": {
        "status": "needs_review",
        "scripted_decisions": 1,
        "effects": 0,
        "intent_events": 0,
        "reason": "context_budget_exceeded",
        "context_budget": {
          "scope": "loop_shared_context",
          "stage": "parameter_selection",
          "unit": "json_characters",
          "limit": 15000,
          "observed": 15001
        }
      },
      "result": {
        "status": "needs_review",
        "reason": "context_budget_exceeded",
        "context_budget": {
          "scope": "loop_shared_context",
          "stage": "parameter_selection",
          "unit": "json_characters",
          "limit": 15000,
          "observed": 15001
        },
        "failure_metadata_correct": true,
        "success_metadata_correct": true,
        "behavior_correct": true,
        "scripted_decisions": 1,
        "effects": 0,
        "intent_events": 0,
        "native_inference_attempts": 0,
        "request_context_sizes": [
          {
            "stage": "tool_selection",
            "characters": 14676,
            "utf8_bytes": 14676
          }
        ],
        "returned_call_telemetry": [
          {
            "stage": "tool_selection",
            "context_characters": 14676,
            "context_character_limit": 15000,
            "context_bytes": 14676
          }
        ],
        "terminal_marker_exposures": 0,
        "reported_scripted_usage_complete": true
      },
      "correct": true,
      "whole_operation_seconds": 0.0016826250066515058,
      "returned_context_bytes": 658,
      "actual_model_usage": {
        "requests": 0,
        "models": {},
        "input_tokens": 0,
        "output_tokens": 0,
        "usage_complete": true
      },
      "arm": "candidate",
      "repeat": 2
    },
    {
      "case": "ascii_parameter_over",
      "runtime_version": "0.1.8",
      "runtime_loaded_from_requested_source": true,
      "fixture": {
        "character": "A",
        "target": "parameter_selection",
        "characters": 15001,
        "marker_prefix_characters": 24,
        "payload_character_repetitions": 14538,
        "frozen_monotonic": 100.0,
        "max_steps": 2,
        "timeout_seconds": 120
      },
      "projected_context_sizes": {
        "context_projection": {
          "characters": 14640,
          "utf8_bytes": 14640
        },
        "tool_selection": {
          "characters": 14676,
          "utf8_bytes": 14676
        },
        "parameter_selection": {
          "characters": 15001,
          "utf8_bytes": 15001
        }
      },
      "expected": {
        "status": "needs_review",
        "scripted_decisions": 1,
        "effects": 0,
        "intent_events": 0,
        "reason": "context_budget_exceeded",
        "context_budget": {
          "scope": "loop_shared_context",
          "stage": "parameter_selection",
          "unit": "json_characters",
          "limit": 15000,
          "observed": 15001
        }
      },
      "result": {
        "status": "needs_review",
        "reason": "context_budget_exceeded",
        "context_budget": null,
        "failure_metadata_correct": false,
        "success_metadata_correct": false,
        "behavior_correct": true,
        "scripted_decisions": 1,
        "effects": 0,
        "intent_events": 0,
        "native_inference_attempts": 0,
        "request_context_sizes": [
          {
            "stage": "tool_selection",
            "characters": 14676,
            "utf8_bytes": 14676
          }
        ],
        "returned_call_telemetry": [
          {
            "stage": "tool_selection",
            "context_characters": null,
            "context_character_limit": null,
            "context_bytes": 14676
          }
        ],
        "terminal_marker_exposures": 0,
        "reported_scripted_usage_complete": true
      },
      "correct": false,
      "whole_operation_seconds": 0.0014749170077266172,
      "returned_context_bytes": 449,
      "actual_model_usage": {
        "requests": 0,
        "models": {},
        "input_tokens": 0,
        "output_tokens": 0,
        "usage_complete": true
      },
      "arm": "baseline",
      "repeat": 2
    },
    {
      "case": "ascii_parameter_over",
      "runtime_version": "0.1.8",
      "runtime_loaded_from_requested_source": true,
      "fixture": {
        "character": "A",
        "target": "parameter_selection",
        "characters": 15001,
        "marker_prefix_characters": 24,
        "payload_character_repetitions": 14538,
        "frozen_monotonic": 100.0,
        "max_steps": 2,
        "timeout_seconds": 120
      },
      "projected_context_sizes": {
        "context_projection": {
          "characters": 14640,
          "utf8_bytes": 14640
        },
        "tool_selection": {
          "characters": 14676,
          "utf8_bytes": 14676
        },
        "parameter_selection": {
          "characters": 15001,
          "utf8_bytes": 15001
        }
      },
      "expected": {
        "status": "needs_review",
        "scripted_decisions": 1,
        "effects": 0,
        "intent_events": 0,
        "reason": "context_budget_exceeded",
        "context_budget": {
          "scope": "loop_shared_context",
          "stage": "parameter_selection",
          "unit": "json_characters",
          "limit": 15000,
          "observed": 15001
        }
      },
      "result": {
        "status": "needs_review",
        "reason": "context_budget_exceeded",
        "context_budget": null,
        "failure_metadata_correct": false,
        "success_metadata_correct": false,
        "behavior_correct": true,
        "scripted_decisions": 1,
        "effects": 0,
        "intent_events": 0,
        "native_inference_attempts": 0,
        "request_context_sizes": [
          {
            "stage": "tool_selection",
            "characters": 14676,
            "utf8_bytes": 14676
          }
        ],
        "returned_call_telemetry": [
          {
            "stage": "tool_selection",
            "context_characters": null,
            "context_character_limit": null,
            "context_bytes": 14676
          }
        ],
        "terminal_marker_exposures": 0,
        "reported_scripted_usage_complete": true
      },
      "correct": false,
      "whole_operation_seconds": 0.001348834004602395,
      "returned_context_bytes": 449,
      "actual_model_usage": {
        "requests": 0,
        "models": {},
        "input_tokens": 0,
        "output_tokens": 0,
        "usage_complete": true
      },
      "arm": "baseline",
      "repeat": 3
    },
    {
      "case": "ascii_parameter_over",
      "runtime_version": "0.1.9",
      "runtime_loaded_from_requested_source": true,
      "fixture": {
        "character": "A",
        "target": "parameter_selection",
        "characters": 15001,
        "marker_prefix_characters": 24,
        "payload_character_repetitions": 14538,
        "frozen_monotonic": 100.0,
        "max_steps": 2,
        "timeout_seconds": 120
      },
      "projected_context_sizes": {
        "context_projection": {
          "characters": 14640,
          "utf8_bytes": 14640
        },
        "tool_selection": {
          "characters": 14676,
          "utf8_bytes": 14676
        },
        "parameter_selection": {
          "characters": 15001,
          "utf8_bytes": 15001
        }
      },
      "expected": {
        "status": "needs_review",
        "scripted_decisions": 1,
        "effects": 0,
        "intent_events": 0,
        "reason": "context_budget_exceeded",
        "context_budget": {
          "scope": "loop_shared_context",
          "stage": "parameter_selection",
          "unit": "json_characters",
          "limit": 15000,
          "observed": 15001
        }
      },
      "result": {
        "status": "needs_review",
        "reason": "context_budget_exceeded",
        "context_budget": {
          "scope": "loop_shared_context",
          "stage": "parameter_selection",
          "unit": "json_characters",
          "limit": 15000,
          "observed": 15001
        },
        "failure_metadata_correct": true,
        "success_metadata_correct": true,
        "behavior_correct": true,
        "scripted_decisions": 1,
        "effects": 0,
        "intent_events": 0,
        "native_inference_attempts": 0,
        "request_context_sizes": [
          {
            "stage": "tool_selection",
            "characters": 14676,
            "utf8_bytes": 14676
          }
        ],
        "returned_call_telemetry": [
          {
            "stage": "tool_selection",
            "context_characters": 14676,
            "context_character_limit": 15000,
            "context_bytes": 14676
          }
        ],
        "terminal_marker_exposures": 0,
        "reported_scripted_usage_complete": true
      },
      "correct": true,
      "whole_operation_seconds": 0.0015137499867705628,
      "returned_context_bytes": 658,
      "actual_model_usage": {
        "requests": 0,
        "models": {},
        "input_tokens": 0,
        "output_tokens": 0,
        "usage_complete": true
      },
      "arm": "candidate",
      "repeat": 3
    },
    {
      "case": "chinese_parameter_exact",
      "runtime_version": "0.1.8",
      "runtime_loaded_from_requested_source": true,
      "fixture": {
        "character": "\u4e2d",
        "target": "parameter_selection",
        "characters": 15000,
        "marker_prefix_characters": 24,
        "payload_character_repetitions": 14537,
        "frozen_monotonic": 100.0,
        "max_steps": 2,
        "timeout_seconds": 120
      },
      "projected_context_sizes": {
        "context_projection": {
          "characters": 14639,
          "utf8_bytes": 43713
        },
        "tool_selection": {
          "characters": 14675,
          "utf8_bytes": 43749
        },
        "parameter_selection": {
          "characters": 15000,
          "utf8_bytes": 44074
        }
      },
      "expected": {
        "status": "done",
        "scripted_decisions": 2,
        "effects": 1,
        "intent_events": 1,
        "reason": null,
        "context_budget": null
      },
      "result": {
        "status": "done",
        "reason": null,
        "context_budget": null,
        "failure_metadata_correct": true,
        "success_metadata_correct": false,
        "behavior_correct": true,
        "scripted_decisions": 2,
        "effects": 1,
        "intent_events": 1,
        "native_inference_attempts": 0,
        "request_context_sizes": [
          {
            "stage": "tool_selection",
            "characters": 14675,
            "utf8_bytes": 43749
          },
          {
            "stage": "parameter_selection",
            "characters": 15000,
            "utf8_bytes": 44074
          }
        ],
        "returned_call_telemetry": [
          {
            "stage": "tool_selection",
            "context_characters": null,
            "context_character_limit": null,
            "context_bytes": 43749
          },
          {
            "stage": "parameter_selection",
            "context_characters": null,
            "context_character_limit": null,
            "context_bytes": 44074
          }
        ],
        "terminal_marker_exposures": 0,
        "reported_scripted_usage_complete": true
      },
      "correct": false,
      "whole_operation_seconds": 0.002477084010024555,
      "returned_context_bytes": 645,
      "actual_model_usage": {
        "requests": 0,
        "models": {},
        "input_tokens": 0,
        "output_tokens": 0,
        "usage_complete": true
      },
      "arm": "baseline",
      "repeat": 1
    },
    {
      "case": "chinese_parameter_exact",
      "runtime_version": "0.1.9",
      "runtime_loaded_from_requested_source": true,
      "fixture": {
        "character": "\u4e2d",
        "target": "parameter_selection",
        "characters": 15000,
        "marker_prefix_characters": 24,
        "payload_character_repetitions": 14537,
        "frozen_monotonic": 100.0,
        "max_steps": 2,
        "timeout_seconds": 120
      },
      "projected_context_sizes": {
        "context_projection": {
          "characters": 14639,
          "utf8_bytes": 43713
        },
        "tool_selection": {
          "characters": 14675,
          "utf8_bytes": 43749
        },
        "parameter_selection": {
          "characters": 15000,
          "utf8_bytes": 44074
        }
      },
      "expected": {
        "status": "done",
        "scripted_decisions": 2,
        "effects": 1,
        "intent_events": 1,
        "reason": null,
        "context_budget": null
      },
      "result": {
        "status": "done",
        "reason": null,
        "context_budget": null,
        "failure_metadata_correct": true,
        "success_metadata_correct": true,
        "behavior_correct": true,
        "scripted_decisions": 2,
        "effects": 1,
        "intent_events": 1,
        "native_inference_attempts": 0,
        "request_context_sizes": [
          {
            "stage": "tool_selection",
            "characters": 14675,
            "utf8_bytes": 43749
          },
          {
            "stage": "parameter_selection",
            "characters": 15000,
            "utf8_bytes": 44074
          }
        ],
        "returned_call_telemetry": [
          {
            "stage": "tool_selection",
            "context_characters": 14675,
            "context_character_limit": 15000,
            "context_bytes": 43749
          },
          {
            "stage": "parameter_selection",
            "context_characters": 15000,
            "context_character_limit": 15000,
            "context_bytes": 44074
          }
        ],
        "terminal_marker_exposures": 0,
        "reported_scripted_usage_complete": true
      },
      "correct": true,
      "whole_operation_seconds": 0.0020680839952547103,
      "returned_context_bytes": 771,
      "actual_model_usage": {
        "requests": 0,
        "models": {},
        "input_tokens": 0,
        "output_tokens": 0,
        "usage_complete": true
      },
      "arm": "candidate",
      "repeat": 1
    },
    {
      "case": "chinese_parameter_exact",
      "runtime_version": "0.1.9",
      "runtime_loaded_from_requested_source": true,
      "fixture": {
        "character": "\u4e2d",
        "target": "parameter_selection",
        "characters": 15000,
        "marker_prefix_characters": 24,
        "payload_character_repetitions": 14537,
        "frozen_monotonic": 100.0,
        "max_steps": 2,
        "timeout_seconds": 120
      },
      "projected_context_sizes": {
        "context_projection": {
          "characters": 14639,
          "utf8_bytes": 43713
        },
        "tool_selection": {
          "characters": 14675,
          "utf8_bytes": 43749
        },
        "parameter_selection": {
          "characters": 15000,
          "utf8_bytes": 44074
        }
      },
      "expected": {
        "status": "done",
        "scripted_decisions": 2,
        "effects": 1,
        "intent_events": 1,
        "reason": null,
        "context_budget": null
      },
      "result": {
        "status": "done",
        "reason": null,
        "context_budget": null,
        "failure_metadata_correct": true,
        "success_metadata_correct": true,
        "behavior_correct": true,
        "scripted_decisions": 2,
        "effects": 1,
        "intent_events": 1,
        "native_inference_attempts": 0,
        "request_context_sizes": [
          {
            "stage": "tool_selection",
            "characters": 14675,
            "utf8_bytes": 43749
          },
          {
            "stage": "parameter_selection",
            "characters": 15000,
            "utf8_bytes": 44074
          }
        ],
        "returned_call_telemetry": [
          {
            "stage": "tool_selection",
            "context_characters": 14675,
            "context_character_limit": 15000,
            "context_bytes": 43749
          },
          {
            "stage": "parameter_selection",
            "context_characters": 15000,
            "context_character_limit": 15000,
            "context_bytes": 44074
          }
        ],
        "terminal_marker_exposures": 0,
        "reported_scripted_usage_complete": true
      },
      "correct": true,
      "whole_operation_seconds": 0.0022466670052381232,
      "returned_context_bytes": 771,
      "actual_model_usage": {
        "requests": 0,
        "models": {},
        "input_tokens": 0,
        "output_tokens": 0,
        "usage_complete": true
      },
      "arm": "candidate",
      "repeat": 2
    },
    {
      "case": "chinese_parameter_exact",
      "runtime_version": "0.1.8",
      "runtime_loaded_from_requested_source": true,
      "fixture": {
        "character": "\u4e2d",
        "target": "parameter_selection",
        "characters": 15000,
        "marker_prefix_characters": 24,
        "payload_character_repetitions": 14537,
        "frozen_monotonic": 100.0,
        "max_steps": 2,
        "timeout_seconds": 120
      },
      "projected_context_sizes": {
        "context_projection": {
          "characters": 14639,
          "utf8_bytes": 43713
        },
        "tool_selection": {
          "characters": 14675,
          "utf8_bytes": 43749
        },
        "parameter_selection": {
          "characters": 15000,
          "utf8_bytes": 44074
        }
      },
      "expected": {
        "status": "done",
        "scripted_decisions": 2,
        "effects": 1,
        "intent_events": 1,
        "reason": null,
        "context_budget": null
      },
      "result": {
        "status": "done",
        "reason": null,
        "context_budget": null,
        "failure_metadata_correct": true,
        "success_metadata_correct": false,
        "behavior_correct": true,
        "scripted_decisions": 2,
        "effects": 1,
        "intent_events": 1,
        "native_inference_attempts": 0,
        "request_context_sizes": [
          {
            "stage": "tool_selection",
            "characters": 14675,
            "utf8_bytes": 43749
          },
          {
            "stage": "parameter_selection",
            "characters": 15000,
            "utf8_bytes": 44074
          }
        ],
        "returned_call_telemetry": [
          {
            "stage": "tool_selection",
            "context_characters": null,
            "context_character_limit": null,
            "context_bytes": 43749
          },
          {
            "stage": "parameter_selection",
            "context_characters": null,
            "context_character_limit": null,
            "context_bytes": 44074
          }
        ],
        "terminal_marker_exposures": 0,
        "reported_scripted_usage_complete": true
      },
      "correct": false,
      "whole_operation_seconds": 0.0022346659970935434,
      "returned_context_bytes": 645,
      "actual_model_usage": {
        "requests": 0,
        "models": {},
        "input_tokens": 0,
        "output_tokens": 0,
        "usage_complete": true
      },
      "arm": "baseline",
      "repeat": 2
    },
    {
      "case": "chinese_parameter_exact",
      "runtime_version": "0.1.8",
      "runtime_loaded_from_requested_source": true,
      "fixture": {
        "character": "\u4e2d",
        "target": "parameter_selection",
        "characters": 15000,
        "marker_prefix_characters": 24,
        "payload_character_repetitions": 14537,
        "frozen_monotonic": 100.0,
        "max_steps": 2,
        "timeout_seconds": 120
      },
      "projected_context_sizes": {
        "context_projection": {
          "characters": 14639,
          "utf8_bytes": 43713
        },
        "tool_selection": {
          "characters": 14675,
          "utf8_bytes": 43749
        },
        "parameter_selection": {
          "characters": 15000,
          "utf8_bytes": 44074
        }
      },
      "expected": {
        "status": "done",
        "scripted_decisions": 2,
        "effects": 1,
        "intent_events": 1,
        "reason": null,
        "context_budget": null
      },
      "result": {
        "status": "done",
        "reason": null,
        "context_budget": null,
        "failure_metadata_correct": true,
        "success_metadata_correct": false,
        "behavior_correct": true,
        "scripted_decisions": 2,
        "effects": 1,
        "intent_events": 1,
        "native_inference_attempts": 0,
        "request_context_sizes": [
          {
            "stage": "tool_selection",
            "characters": 14675,
            "utf8_bytes": 43749
          },
          {
            "stage": "parameter_selection",
            "characters": 15000,
            "utf8_bytes": 44074
          }
        ],
        "returned_call_telemetry": [
          {
            "stage": "tool_selection",
            "context_characters": null,
            "context_character_limit": null,
            "context_bytes": 43749
          },
          {
            "stage": "parameter_selection",
            "context_characters": null,
            "context_character_limit": null,
            "context_bytes": 44074
          }
        ],
        "terminal_marker_exposures": 0,
        "reported_scripted_usage_complete": true
      },
      "correct": false,
      "whole_operation_seconds": 0.0022021249897079542,
      "returned_context_bytes": 645,
      "actual_model_usage": {
        "requests": 0,
        "models": {},
        "input_tokens": 0,
        "output_tokens": 0,
        "usage_complete": true
      },
      "arm": "baseline",
      "repeat": 3
    },
    {
      "case": "chinese_parameter_exact",
      "runtime_version": "0.1.9",
      "runtime_loaded_from_requested_source": true,
      "fixture": {
        "character": "\u4e2d",
        "target": "parameter_selection",
        "characters": 15000,
        "marker_prefix_characters": 24,
        "payload_character_repetitions": 14537,
        "frozen_monotonic": 100.0,
        "max_steps": 2,
        "timeout_seconds": 120
      },
      "projected_context_sizes": {
        "context_projection": {
          "characters": 14639,
          "utf8_bytes": 43713
        },
        "tool_selection": {
          "characters": 14675,
          "utf8_bytes": 43749
        },
        "parameter_selection": {
          "characters": 15000,
          "utf8_bytes": 44074
        }
      },
      "expected": {
        "status": "done",
        "scripted_decisions": 2,
        "effects": 1,
        "intent_events": 1,
        "reason": null,
        "context_budget": null
      },
      "result": {
        "status": "done",
        "reason": null,
        "context_budget": null,
        "failure_metadata_correct": true,
        "success_metadata_correct": true,
        "behavior_correct": true,
        "scripted_decisions": 2,
        "effects": 1,
        "intent_events": 1,
        "native_inference_attempts": 0,
        "request_context_sizes": [
          {
            "stage": "tool_selection",
            "characters": 14675,
            "utf8_bytes": 43749
          },
          {
            "stage": "parameter_selection",
            "characters": 15000,
            "utf8_bytes": 44074
          }
        ],
        "returned_call_telemetry": [
          {
            "stage": "tool_selection",
            "context_characters": 14675,
            "context_character_limit": 15000,
            "context_bytes": 43749
          },
          {
            "stage": "parameter_selection",
            "context_characters": 15000,
            "context_character_limit": 15000,
            "context_bytes": 44074
          }
        ],
        "terminal_marker_exposures": 0,
        "reported_scripted_usage_complete": true
      },
      "correct": true,
      "whole_operation_seconds": 0.002064875006908551,
      "returned_context_bytes": 771,
      "actual_model_usage": {
        "requests": 0,
        "models": {},
        "input_tokens": 0,
        "output_tokens": 0,
        "usage_complete": true
      },
      "arm": "candidate",
      "repeat": 3
    },
    {
      "case": "direct_ascii_over",
      "runtime_version": "0.1.8",
      "runtime_loaded_from_requested_source": true,
      "fixture": {
        "character": "A",
        "target": "direct_decide",
        "characters": 15001,
        "marker_prefix_characters": 24,
        "payload_character_repetitions": 14962,
        "frozen_monotonic": 100.0,
        "max_steps": 2,
        "timeout_seconds": 120
      },
      "projected_context_sizes": {
        "direct_decide": {
          "characters": 15001,
          "utf8_bytes": 15001
        }
      },
      "expected": {
        "status": "admitted",
        "scripted_decisions": 1,
        "effects": 0,
        "intent_events": 0,
        "reason": null,
        "context_budget": null
      },
      "result": {
        "status": "admitted",
        "reason": null,
        "context_budget": null,
        "failure_metadata_correct": true,
        "success_metadata_correct": true,
        "behavior_correct": true,
        "scripted_decisions": 1,
        "effects": 0,
        "intent_events": 0,
        "native_inference_attempts": 0,
        "request_context_sizes": [
          {
            "stage": "direct_decide",
            "characters": 15001,
            "utf8_bytes": 15001
          }
        ],
        "returned_call_telemetry": [],
        "terminal_marker_exposures": 0,
        "reported_scripted_usage_complete": true
      },
      "correct": true,
      "whole_operation_seconds": 0.000888833004864864,
      "returned_context_bytes": 246,
      "actual_model_usage": {
        "requests": 0,
        "models": {},
        "input_tokens": 0,
        "output_tokens": 0,
        "usage_complete": true
      },
      "arm": "baseline",
      "repeat": 1
    },
    {
      "case": "direct_ascii_over",
      "runtime_version": "0.1.9",
      "runtime_loaded_from_requested_source": true,
      "fixture": {
        "character": "A",
        "target": "direct_decide",
        "characters": 15001,
        "marker_prefix_characters": 24,
        "payload_character_repetitions": 14962,
        "frozen_monotonic": 100.0,
        "max_steps": 2,
        "timeout_seconds": 120
      },
      "projected_context_sizes": {
        "direct_decide": {
          "characters": 15001,
          "utf8_bytes": 15001
        }
      },
      "expected": {
        "status": "admitted",
        "scripted_decisions": 1,
        "effects": 0,
        "intent_events": 0,
        "reason": null,
        "context_budget": null
      },
      "result": {
        "status": "admitted",
        "reason": null,
        "context_budget": null,
        "failure_metadata_correct": true,
        "success_metadata_correct": true,
        "behavior_correct": true,
        "scripted_decisions": 1,
        "effects": 0,
        "intent_events": 0,
        "native_inference_attempts": 0,
        "request_context_sizes": [
          {
            "stage": "direct_decide",
            "characters": 15001,
            "utf8_bytes": 15001
          }
        ],
        "returned_call_telemetry": [],
        "terminal_marker_exposures": 0,
        "reported_scripted_usage_complete": true
      },
      "correct": true,
      "whole_operation_seconds": 0.0010096250043716282,
      "returned_context_bytes": 246,
      "actual_model_usage": {
        "requests": 0,
        "models": {},
        "input_tokens": 0,
        "output_tokens": 0,
        "usage_complete": true
      },
      "arm": "candidate",
      "repeat": 1
    },
    {
      "case": "direct_ascii_over",
      "runtime_version": "0.1.9",
      "runtime_loaded_from_requested_source": true,
      "fixture": {
        "character": "A",
        "target": "direct_decide",
        "characters": 15001,
        "marker_prefix_characters": 24,
        "payload_character_repetitions": 14962,
        "frozen_monotonic": 100.0,
        "max_steps": 2,
        "timeout_seconds": 120
      },
      "projected_context_sizes": {
        "direct_decide": {
          "characters": 15001,
          "utf8_bytes": 15001
        }
      },
      "expected": {
        "status": "admitted",
        "scripted_decisions": 1,
        "effects": 0,
        "intent_events": 0,
        "reason": null,
        "context_budget": null
      },
      "result": {
        "status": "admitted",
        "reason": null,
        "context_budget": null,
        "failure_metadata_correct": true,
        "success_metadata_correct": true,
        "behavior_correct": true,
        "scripted_decisions": 1,
        "effects": 0,
        "intent_events": 0,
        "native_inference_attempts": 0,
        "request_context_sizes": [
          {
            "stage": "direct_decide",
            "characters": 15001,
            "utf8_bytes": 15001
          }
        ],
        "returned_call_telemetry": [],
        "terminal_marker_exposures": 0,
        "reported_scripted_usage_complete": true
      },
      "correct": true,
      "whole_operation_seconds": 0.0008822920062812045,
      "returned_context_bytes": 246,
      "actual_model_usage": {
        "requests": 0,
        "models": {},
        "input_tokens": 0,
        "output_tokens": 0,
        "usage_complete": true
      },
      "arm": "candidate",
      "repeat": 2
    },
    {
      "case": "direct_ascii_over",
      "runtime_version": "0.1.8",
      "runtime_loaded_from_requested_source": true,
      "fixture": {
        "character": "A",
        "target": "direct_decide",
        "characters": 15001,
        "marker_prefix_characters": 24,
        "payload_character_repetitions": 14962,
        "frozen_monotonic": 100.0,
        "max_steps": 2,
        "timeout_seconds": 120
      },
      "projected_context_sizes": {
        "direct_decide": {
          "characters": 15001,
          "utf8_bytes": 15001
        }
      },
      "expected": {
        "status": "admitted",
        "scripted_decisions": 1,
        "effects": 0,
        "intent_events": 0,
        "reason": null,
        "context_budget": null
      },
      "result": {
        "status": "admitted",
        "reason": null,
        "context_budget": null,
        "failure_metadata_correct": true,
        "success_metadata_correct": true,
        "behavior_correct": true,
        "scripted_decisions": 1,
        "effects": 0,
        "intent_events": 0,
        "native_inference_attempts": 0,
        "request_context_sizes": [
          {
            "stage": "direct_decide",
            "characters": 15001,
            "utf8_bytes": 15001
          }
        ],
        "returned_call_telemetry": [],
        "terminal_marker_exposures": 0,
        "reported_scripted_usage_complete": true
      },
      "correct": true,
      "whole_operation_seconds": 0.0007155409984989092,
      "returned_context_bytes": 246,
      "actual_model_usage": {
        "requests": 0,
        "models": {},
        "input_tokens": 0,
        "output_tokens": 0,
        "usage_complete": true
      },
      "arm": "baseline",
      "repeat": 2
    },
    {
      "case": "direct_ascii_over",
      "runtime_version": "0.1.8",
      "runtime_loaded_from_requested_source": true,
      "fixture": {
        "character": "A",
        "target": "direct_decide",
        "characters": 15001,
        "marker_prefix_characters": 24,
        "payload_character_repetitions": 14962,
        "frozen_monotonic": 100.0,
        "max_steps": 2,
        "timeout_seconds": 120
      },
      "projected_context_sizes": {
        "direct_decide": {
          "characters": 15001,
          "utf8_bytes": 15001
        }
      },
      "expected": {
        "status": "admitted",
        "scripted_decisions": 1,
        "effects": 0,
        "intent_events": 0,
        "reason": null,
        "context_budget": null
      },
      "result": {
        "status": "admitted",
        "reason": null,
        "context_budget": null,
        "failure_metadata_correct": true,
        "success_metadata_correct": true,
        "behavior_correct": true,
        "scripted_decisions": 1,
        "effects": 0,
        "intent_events": 0,
        "native_inference_attempts": 0,
        "request_context_sizes": [
          {
            "stage": "direct_decide",
            "characters": 15001,
            "utf8_bytes": 15001
          }
        ],
        "returned_call_telemetry": [],
        "terminal_marker_exposures": 0,
        "reported_scripted_usage_complete": true
      },
      "correct": true,
      "whole_operation_seconds": 0.0009514170087641105,
      "returned_context_bytes": 246,
      "actual_model_usage": {
        "requests": 0,
        "models": {},
        "input_tokens": 0,
        "output_tokens": 0,
        "usage_complete": true
      },
      "arm": "baseline",
      "repeat": 3
    },
    {
      "case": "direct_ascii_over",
      "runtime_version": "0.1.9",
      "runtime_loaded_from_requested_source": true,
      "fixture": {
        "character": "A",
        "target": "direct_decide",
        "characters": 15001,
        "marker_prefix_characters": 24,
        "payload_character_repetitions": 14962,
        "frozen_monotonic": 100.0,
        "max_steps": 2,
        "timeout_seconds": 120
      },
      "projected_context_sizes": {
        "direct_decide": {
          "characters": 15001,
          "utf8_bytes": 15001
        }
      },
      "expected": {
        "status": "admitted",
        "scripted_decisions": 1,
        "effects": 0,
        "intent_events": 0,
        "reason": null,
        "context_budget": null
      },
      "result": {
        "status": "admitted",
        "reason": null,
        "context_budget": null,
        "failure_metadata_correct": true,
        "success_metadata_correct": true,
        "behavior_correct": true,
        "scripted_decisions": 1,
        "effects": 0,
        "intent_events": 0,
        "native_inference_attempts": 0,
        "request_context_sizes": [
          {
            "stage": "direct_decide",
            "characters": 15001,
            "utf8_bytes": 15001
          }
        ],
        "returned_call_telemetry": [],
        "terminal_marker_exposures": 0,
        "reported_scripted_usage_complete": true
      },
      "correct": true,
      "whole_operation_seconds": 0.0009281250095227733,
      "returned_context_bytes": 246,
      "actual_model_usage": {
        "requests": 0,
        "models": {},
        "input_tokens": 0,
        "output_tokens": 0,
        "usage_complete": true
      },
      "arm": "candidate",
      "repeat": 3
    },
    {
      "case": "direct_chinese_over",
      "runtime_version": "0.1.8",
      "runtime_loaded_from_requested_source": true,
      "fixture": {
        "character": "\u4e2d",
        "target": "direct_decide",
        "characters": 15001,
        "marker_prefix_characters": 24,
        "payload_character_repetitions": 14962,
        "frozen_monotonic": 100.0,
        "max_steps": 2,
        "timeout_seconds": 120
      },
      "projected_context_sizes": {
        "direct_decide": {
          "characters": 15001,
          "utf8_bytes": 44925
        }
      },
      "expected": {
        "status": "admitted",
        "scripted_decisions": 1,
        "effects": 0,
        "intent_events": 0,
        "reason": null,
        "context_budget": null
      },
      "result": {
        "status": "admitted",
        "reason": null,
        "context_budget": null,
        "failure_metadata_correct": true,
        "success_metadata_correct": true,
        "behavior_correct": true,
        "scripted_decisions": 1,
        "effects": 0,
        "intent_events": 0,
        "native_inference_attempts": 0,
        "request_context_sizes": [
          {
            "stage": "direct_decide",
            "characters": 15001,
            "utf8_bytes": 44925
          }
        ],
        "returned_call_telemetry": [],
        "terminal_marker_exposures": 0,
        "reported_scripted_usage_complete": true
      },
      "correct": true,
      "whole_operation_seconds": 0.0009609580010874197,
      "returned_context_bytes": 246,
      "actual_model_usage": {
        "requests": 0,
        "models": {},
        "input_tokens": 0,
        "output_tokens": 0,
        "usage_complete": true
      },
      "arm": "baseline",
      "repeat": 1
    },
    {
      "case": "direct_chinese_over",
      "runtime_version": "0.1.9",
      "runtime_loaded_from_requested_source": true,
      "fixture": {
        "character": "\u4e2d",
        "target": "direct_decide",
        "characters": 15001,
        "marker_prefix_characters": 24,
        "payload_character_repetitions": 14962,
        "frozen_monotonic": 100.0,
        "max_steps": 2,
        "timeout_seconds": 120
      },
      "projected_context_sizes": {
        "direct_decide": {
          "characters": 15001,
          "utf8_bytes": 44925
        }
      },
      "expected": {
        "status": "admitted",
        "scripted_decisions": 1,
        "effects": 0,
        "intent_events": 0,
        "reason": null,
        "context_budget": null
      },
      "result": {
        "status": "admitted",
        "reason": null,
        "context_budget": null,
        "failure_metadata_correct": true,
        "success_metadata_correct": true,
        "behavior_correct": true,
        "scripted_decisions": 1,
        "effects": 0,
        "intent_events": 0,
        "native_inference_attempts": 0,
        "request_context_sizes": [
          {
            "stage": "direct_decide",
            "characters": 15001,
            "utf8_bytes": 44925
          }
        ],
        "returned_call_telemetry": [],
        "terminal_marker_exposures": 0,
        "reported_scripted_usage_complete": true
      },
      "correct": true,
      "whole_operation_seconds": 0.0011525419977260754,
      "returned_context_bytes": 246,
      "actual_model_usage": {
        "requests": 0,
        "models": {},
        "input_tokens": 0,
        "output_tokens": 0,
        "usage_complete": true
      },
      "arm": "candidate",
      "repeat": 1
    },
    {
      "case": "direct_chinese_over",
      "runtime_version": "0.1.9",
      "runtime_loaded_from_requested_source": true,
      "fixture": {
        "character": "\u4e2d",
        "target": "direct_decide",
        "characters": 15001,
        "marker_prefix_characters": 24,
        "payload_character_repetitions": 14962,
        "frozen_monotonic": 100.0,
        "max_steps": 2,
        "timeout_seconds": 120
      },
      "projected_context_sizes": {
        "direct_decide": {
          "characters": 15001,
          "utf8_bytes": 44925
        }
      },
      "expected": {
        "status": "admitted",
        "scripted_decisions": 1,
        "effects": 0,
        "intent_events": 0,
        "reason": null,
        "context_budget": null
      },
      "result": {
        "status": "admitted",
        "reason": null,
        "context_budget": null,
        "failure_metadata_correct": true,
        "success_metadata_correct": true,
        "behavior_correct": true,
        "scripted_decisions": 1,
        "effects": 0,
        "intent_events": 0,
        "native_inference_attempts": 0,
        "request_context_sizes": [
          {
            "stage": "direct_decide",
            "characters": 15001,
            "utf8_bytes": 44925
          }
        ],
        "returned_call_telemetry": [],
        "terminal_marker_exposures": 0,
        "reported_scripted_usage_complete": true
      },
      "correct": true,
      "whole_operation_seconds": 0.0009513339900877327,
      "returned_context_bytes": 246,
      "actual_model_usage": {
        "requests": 0,
        "models": {},
        "input_tokens": 0,
        "output_tokens": 0,
        "usage_complete": true
      },
      "arm": "candidate",
      "repeat": 2
    },
    {
      "case": "direct_chinese_over",
      "runtime_version": "0.1.8",
      "runtime_loaded_from_requested_source": true,
      "fixture": {
        "character": "\u4e2d",
        "target": "direct_decide",
        "characters": 15001,
        "marker_prefix_characters": 24,
        "payload_character_repetitions": 14962,
        "frozen_monotonic": 100.0,
        "max_steps": 2,
        "timeout_seconds": 120
      },
      "projected_context_sizes": {
        "direct_decide": {
          "characters": 15001,
          "utf8_bytes": 44925
        }
      },
      "expected": {
        "status": "admitted",
        "scripted_decisions": 1,
        "effects": 0,
        "intent_events": 0,
        "reason": null,
        "context_budget": null
      },
      "result": {
        "status": "admitted",
        "reason": null,
        "context_budget": null,
        "failure_metadata_correct": true,
        "success_metadata_correct": true,
        "behavior_correct": true,
        "scripted_decisions": 1,
        "effects": 0,
        "intent_events": 0,
        "native_inference_attempts": 0,
        "request_context_sizes": [
          {
            "stage": "direct_decide",
            "characters": 15001,
            "utf8_bytes": 44925
          }
        ],
        "returned_call_telemetry": [],
        "terminal_marker_exposures": 0,
        "reported_scripted_usage_complete": true
      },
      "correct": true,
      "whole_operation_seconds": 0.0009500419982941821,
      "returned_context_bytes": 246,
      "actual_model_usage": {
        "requests": 0,
        "models": {},
        "input_tokens": 0,
        "output_tokens": 0,
        "usage_complete": true
      },
      "arm": "baseline",
      "repeat": 2
    },
    {
      "case": "direct_chinese_over",
      "runtime_version": "0.1.8",
      "runtime_loaded_from_requested_source": true,
      "fixture": {
        "character": "\u4e2d",
        "target": "direct_decide",
        "characters": 15001,
        "marker_prefix_characters": 24,
        "payload_character_repetitions": 14962,
        "frozen_monotonic": 100.0,
        "max_steps": 2,
        "timeout_seconds": 120
      },
      "projected_context_sizes": {
        "direct_decide": {
          "characters": 15001,
          "utf8_bytes": 44925
        }
      },
      "expected": {
        "status": "admitted",
        "scripted_decisions": 1,
        "effects": 0,
        "intent_events": 0,
        "reason": null,
        "context_budget": null
      },
      "result": {
        "status": "admitted",
        "reason": null,
        "context_budget": null,
        "failure_metadata_correct": true,
        "success_metadata_correct": true,
        "behavior_correct": true,
        "scripted_decisions": 1,
        "effects": 0,
        "intent_events": 0,
        "native_inference_attempts": 0,
        "request_context_sizes": [
          {
            "stage": "direct_decide",
            "characters": 15001,
            "utf8_bytes": 44925
          }
        ],
        "returned_call_telemetry": [],
        "terminal_marker_exposures": 0,
        "reported_scripted_usage_complete": true
      },
      "correct": true,
      "whole_operation_seconds": 0.0009688749996712431,
      "returned_context_bytes": 246,
      "actual_model_usage": {
        "requests": 0,
        "models": {},
        "input_tokens": 0,
        "output_tokens": 0,
        "usage_complete": true
      },
      "arm": "baseline",
      "repeat": 3
    },
    {
      "case": "direct_chinese_over",
      "runtime_version": "0.1.9",
      "runtime_loaded_from_requested_source": true,
      "fixture": {
        "character": "\u4e2d",
        "target": "direct_decide",
        "characters": 15001,
        "marker_prefix_characters": 24,
        "payload_character_repetitions": 14962,
        "frozen_monotonic": 100.0,
        "max_steps": 2,
        "timeout_seconds": 120
      },
      "projected_context_sizes": {
        "direct_decide": {
          "characters": 15001,
          "utf8_bytes": 44925
        }
      },
      "expected": {
        "status": "admitted",
        "scripted_decisions": 1,
        "effects": 0,
        "intent_events": 0,
        "reason": null,
        "context_budget": null
      },
      "result": {
        "status": "admitted",
        "reason": null,
        "context_budget": null,
        "failure_metadata_correct": true,
        "success_metadata_correct": true,
        "behavior_correct": true,
        "scripted_decisions": 1,
        "effects": 0,
        "intent_events": 0,
        "native_inference_attempts": 0,
        "request_context_sizes": [
          {
            "stage": "direct_decide",
            "characters": 15001,
            "utf8_bytes": 44925
          }
        ],
        "returned_call_telemetry": [],
        "terminal_marker_exposures": 0,
        "reported_scripted_usage_complete": true
      },
      "correct": true,
      "whole_operation_seconds": 0.0009766660077730194,
      "returned_context_bytes": 246,
      "actual_model_usage": {
        "requests": 0,
        "models": {},
        "input_tokens": 0,
        "output_tokens": 0,
        "usage_complete": true
      },
      "arm": "candidate",
      "repeat": 3
    }
  ]
}